framework,version,device,op_name,kernel_source,mla_dtype,kv_cache_dtype,num_heads,batch_size,isl,tp_size,step,latency
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,1,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,1,0.01048159971833229
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,1,0.01067039966583252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,1,0.010611200332641601
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,1,0.010502400249242783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,1,0.010491199791431427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,1,0.010611200332641601
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,1,0.01048479974269867
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,1,0.01043199971318245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,1,0.010440000146627427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,1,0.010500799864530563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,1,0.010465600341558457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,1,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,1,0.010499200224876404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,3,0.010764800012111664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,3,0.012591999769210816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,3,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,3,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,3,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,3,0.010702399909496308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,3,0.010526400059461594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,3,0.010809600353240967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,3,0.010527999699115753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,3,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,3,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,3,0.010480000078678131
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,3,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,3,0.010859200358390808
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,7,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,7,0.010527999699115753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,7,0.012595200538635254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,7,0.010819199681282043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,7,0.010520000010728836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,7,0.010636799782514573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,7,0.01069599986076355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,7,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,7,0.01050880029797554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,7,0.010576000064611435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,7,0.01053439974784851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,7,0.010623999685049058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,7,0.010571199655532836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,7,0.010531199723482132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,15,0.010583999752998351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,15,0.01183359995484352
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,15,0.01067200005054474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,15,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,15,0.010527999699115753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,15,0.010768000036478043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,15,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,15,0.010577599704265594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,15,0.010694400221109391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,15,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,15,0.009464000165462495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,15,0.010793600231409073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,15,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,15,0.00934239998459816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,31,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,31,0.011523199826478958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,31,0.010838399827480315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,31,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,31,0.010657600313425063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,31,0.010691200196743012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,31,0.010659199953079224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,31,0.01112319976091385
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,31,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,31,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,31,0.010523200035095215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,31,0.009752000123262406
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,31,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,31,0.010545600205659866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,63,0.010623999685049058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,63,0.01077599972486496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,63,0.010628800094127654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,63,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,63,0.010651200264692306
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,63,0.010496000200510025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,63,0.010665600001811982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,63,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,63,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,63,0.010545600205659866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,63,0.010526400059461594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,63,0.01058880016207695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,63,0.010636799782514573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,63,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,127,0.01077599972486496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,127,0.012540799379348756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,127,0.010768000036478043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,127,0.01061279997229576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,127,0.010576000064611435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,127,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,127,0.010793600231409073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,127,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,127,0.01061440035700798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,127,0.010702399909496308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,127,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,127,0.009625600278377533
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,127,0.008950400352478027
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,127,0.01061599999666214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,255,0.010622400045394897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,255,0.010531199723482132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,255,0.010556799918413162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,255,0.012510399520397186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,255,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,255,0.010556799918413162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,255,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,255,0.010524799674749374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,255,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,255,0.01061599999666214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,255,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,255,0.009611199796199798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,255,0.008942399919033051
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,255,0.010505600273609162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,511,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,511,0.012788799405097962
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,511,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,511,0.012625600397586822
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,511,0.012678399682044983
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,511,0.012630400061607362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,511,0.012702399492263794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,511,0.012654399871826172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,511,0.011694400012493134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,511,0.012564800679683685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,511,0.011323200166225433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,511,0.012628799676895142
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,511,0.01234079971909523
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,511,0.01067200005054474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,1023,0.01480959951877594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,1023,0.016659200191497803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,1023,0.016675199568271636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,1023,0.014662399888038635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,1023,0.014686399698257446
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,1023,0.014777599275112152
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,1023,0.014744000136852264
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,1023,0.014569599926471711
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,1023,0.014723199605941772
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,1023,0.014777599275112152
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,1023,0.01271039992570877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,1023,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,1023,0.012831999361515046
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,1023,0.012860800325870513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,2047,0.0165583997964859
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,2047,0.01478559970855713
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,2047,0.01653600037097931
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,2047,0.014572800695896148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,2047,0.014910399913787842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,2047,0.014664000272750855
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,2047,0.014697599411010741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,2047,0.014830400049686433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,2047,0.015559999644756317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,2047,0.014668799936771393
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,2047,0.012913599610328674
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,2047,0.01340160071849823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,2047,0.01465280055999756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,2047,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,4095,0.017233599722385407
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,4095,0.016910399496555328
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,4095,0.01685120016336441
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,4095,0.014643199741840363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,4095,0.014734399318695069
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,4095,0.014859199523925781
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,4095,0.014777599275112152
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,4095,0.014819200336933135
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,4095,0.01478399932384491
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,4095,0.01494240015745163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,4095,0.014612799882888794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,4095,0.012801599502563477
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,4095,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,4095,0.01465280055999756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,8191,0.01865279972553253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,8191,0.016864000260829924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,8191,0.02380799949169159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,8191,0.01672160029411316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,8191,0.015252800285816192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,8191,0.016769599914550782
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,8191,0.01687999963760376
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,8191,0.018918399512767792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,8191,0.016711999475955964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,8191,0.016313600540161132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,8191,0.014779199659824372
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,8191,0.014817599952220917
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,8191,0.014897599816322327
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,8191,0.014854399859905243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,16383,0.02678399980068207
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,16383,0.02808000147342682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,16383,0.020772799849510193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,16383,0.01915840059518814
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,16383,0.019043199717998505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,16383,0.020703999698162077
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,16383,0.019200000166893005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,16383,0.022916799783706664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,16383,0.021547199785709382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,16383,0.018990400433540344
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,16383,0.016891199350357055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,16383,0.01804639995098114
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,16383,0.018884800374507904
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,16383,0.01886080056428909
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,32767,0.03691839873790741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,32767,0.03105599880218506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,32767,0.02911359965801239
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,32767,0.027153599262237548
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,32767,0.027105599641799927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,32767,0.027001601457595826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,32767,0.02778719961643219
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,32767,0.025167998671531678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,32767,0.02396479994058609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,32767,0.023102399706840516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,32767,0.023014399409294128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,32767,0.022924800217151643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,32767,0.022801600396633148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,32767,0.022912000119686127
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,65535,0.04280959963798523
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,65535,0.03499200046062469
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,65535,0.03927040100097656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,65535,0.03425759971141815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,65535,0.03470560014247894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,65535,0.033913600444793704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,65535,0.03611840009689331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,65535,0.03152799904346466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,65535,0.028472000360488893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,65535,0.027224001288414002
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,65535,0.02728480100631714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,65535,0.027185600996017457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,65535,0.02709920108318329
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,65535,0.02744640111923218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,1,131071,0.05781279802322388
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,2,131071,0.0686784029006958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,4,131071,0.053337597846984865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,8,131071,0.05300160050392151
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,16,131071,0.05257120132446289
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,32,131071,0.05426880121231079
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1,1,64,131071,0.055190402269363406
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,1,131071,0.039473599195480345
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,2,131071,0.04024159908294678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,4,131071,0.0330592006444931
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,8,131071,0.0335072010755539
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,16,131071,0.03323040008544922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,32,131071,0.031913599371910094
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1,1,64,131071,0.03293919861316681
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,1,0.011257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,1,0.011483199894428253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,1,0.010732799768447876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,1,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,1,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,1,0.010633599758148194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,1,0.010694400221109391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,1,0.010500799864530563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,1,0.010627199709415436
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,1,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,1,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,1,0.010684800148010255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,1,0.01053439974784851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,1,0.010632000118494033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,3,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,3,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,3,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,3,0.010787200182676315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,3,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,3,0.010652799904346467
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,3,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,3,0.0107744000852108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,3,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,3,0.01069760024547577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,3,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,3,0.010571199655532836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,3,0.010577599704265594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,3,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,7,0.012736000120639801
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,7,0.012627199292182922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,7,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,7,0.010496000200510025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,7,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,7,0.010790400207042694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,7,0.010552000254392624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,7,0.010491199791431427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,7,0.010652799904346467
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,7,0.010636799782514573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,7,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,7,0.010611200332641601
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,7,0.010446400195360184
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,7,0.010552000254392624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,15,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,15,0.010980799794197083
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,15,0.010833600163459777
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,15,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,15,0.010478399693965912
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,15,0.010636799782514573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,15,0.010569600015878677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,15,0.010831999778747558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,15,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,15,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,15,0.01056160032749176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,15,0.010579200088977813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,15,0.010566399991512298
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,15,0.010761599987745285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,31,0.01069599986076355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,31,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,31,0.010732799768447876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,31,0.010761599987745285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,31,0.01071999967098236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,31,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,31,0.010463999956846238
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,31,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,31,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,31,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,31,0.010518400371074677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,31,0.010608000308275222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,31,0.01063840016722679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,31,0.010494399815797806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,63,0.01127839982509613
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,63,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,63,0.010911999642848969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,63,0.010608000308275222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,63,0.011523199826478958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,63,0.01063840016722679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,63,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,63,0.01064639985561371
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,63,0.010812799632549285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,63,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,63,0.010500799864530563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,63,0.010572800040245056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,63,0.010577599704265594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,63,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,127,0.01072480008006096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,127,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,127,0.012678399682044983
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,127,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,127,0.010764800012111664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,127,0.010804799944162368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,127,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,127,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,127,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,127,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,127,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,127,0.010683199763298035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,127,0.010579200088977813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,127,0.01056319996714592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,255,0.010753600299358368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,255,0.010926400125026704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,255,0.010675200074911118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,255,0.010844799876213073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,255,0.01061279997229576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,255,0.010531199723482132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,255,0.010876800119876861
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,255,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,255,0.010622400045394897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,255,0.010768000036478043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,255,0.010553599894046783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,255,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,255,0.010579200088977813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,255,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,511,0.012796799838542938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,511,0.013654400408267976
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,511,0.012904000282287598
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,511,0.012747199833393097
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,511,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,511,0.012758399546146392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,511,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,511,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,511,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,511,0.01273919939994812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,511,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,511,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,511,0.010731200128793717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,511,0.011209599673748016
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,1023,0.014734399318695069
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,1023,0.014860799908638
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,1023,0.014523200690746307
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,1023,0.014692799746990204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,1023,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,1023,0.01279519945383072
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,1023,0.012510399520397186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,1023,0.012759999930858612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,1023,0.012759999930858612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,1023,0.012716799974441528
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,1023,0.012622399628162384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,1023,0.01273760050535202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,1023,0.01260959953069687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,1023,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,2047,0.016864000260829924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,2047,0.01462559998035431
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,2047,0.014779199659824372
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,2047,0.014577600359916686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,2047,0.012740799784660339
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,2047,0.0127920001745224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,2047,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,2047,0.014547200500965118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,2047,0.01284320056438446
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,2047,0.012652799487113953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,2047,0.01284639984369278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,2047,0.012580800056457519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,2047,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,2047,0.012593600153923034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,4095,0.01863519996404648
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,4095,0.01884640008211136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,4095,0.01481119990348816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,4095,0.014616000652313232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,4095,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,4095,0.014635199308395385
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,4095,0.012908799946308136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,4095,0.016680000722408293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,4095,0.016780799627304076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,4095,0.012830400466918945
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,4095,0.01268800050020218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,4095,0.01279360055923462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,4095,0.012809599936008453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,4095,0.012585599720478059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,8191,0.021089600026607515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,8191,0.019126400351524353
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,8191,0.019049599766731262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,8191,0.01684959977865219
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,8191,0.017521600425243377
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,8191,0.01686079949140549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,8191,0.017918400466442108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,8191,0.016899199783802034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,8191,0.01698399931192398
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,8191,0.01675360053777695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,8191,0.01677280068397522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,8191,0.014820800721645355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,8191,0.01675039976835251
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,8191,0.015279999375343323
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,16383,0.032652801275253295
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,16383,0.02285439968109131
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,16383,0.021348799765110015
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,16383,0.02091040015220642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,16383,0.020803199708461763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,16383,0.01881919950246811
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,16383,0.020759999752044678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,16383,0.023340800404548646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,16383,0.020764799416065217
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,16383,0.018697600066661834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,16383,0.018939200043678283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,16383,0.018691200017929076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,16383,0.018822400271892546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,16383,0.016891199350357055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,32767,0.027643200755119324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,32767,0.025148800015449523
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,32767,0.035304000973701476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,32767,0.03681600093841553
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,32767,0.02696000039577484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,32767,0.02507520020008087
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,32767,0.027004799246788024
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,32767,0.03544479906558991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,32767,0.025284799933433532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,32767,0.022969600558280946
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,32767,0.02297919988632202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,32767,0.023244799673557283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,32767,0.02298559993505478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,32767,0.022572800517082214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,65535,0.05634400248527527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,65535,0.05543680191040039
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,65535,0.05338559746742248
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,65535,0.05097439885139465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,65535,0.0539359986782074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,65535,0.05368959903717041
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,65535,0.05427520275115967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,65535,0.035471999645233156
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,65535,0.03319680094718933
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,65535,0.03120799958705902
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,65535,0.031137600541114807
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,65535,0.029337599873542786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,65535,0.029151999950408937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,65535,0.029398399591445922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,1,131071,0.08230400085449219
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,4,131071,0.07542240023612976
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,2,131071,0.08954079747200012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,8,131071,0.07448639869689941
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,16,131071,0.07716799974441528
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,32,131071,0.07656959891319275
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,2,1,64,131071,0.07666239738464356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,1,131071,0.05376480221748352
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,4,131071,0.04766879975795746
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,2,131071,0.05157439708709717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,8,131071,0.046265599131584165
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,32,131071,0.048655998706817624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,16,131071,0.04879199862480164
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,2,1,64,131071,0.04793919920921326
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,1,0.010764800012111664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,1,0.010782399773597717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,1,0.01260959953069687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,1,0.01080159991979599
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,1,0.01079839989542961
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,1,0.010513599961996078
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,1,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,1,0.01055999994277954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,1,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,1,0.010552000254392624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,1,0.010532800108194351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,1,0.010523200035095215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,1,0.010569600015878677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,1,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,3,0.012721599638462066
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,3,0.011179199814796448
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,3,0.010732799768447876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,3,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,3,0.010676799714565277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,3,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,3,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,3,0.010886400192975997
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,3,0.010655999928712846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,3,0.010731200128793717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,3,0.010620799660682679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,3,0.010521599650382995
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,3,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,3,0.010623999685049058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,7,0.012697599828243256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,7,0.011360000073909759
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,7,0.012697599828243256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,7,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,7,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,7,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,7,0.010822399705648422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,7,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,7,0.010548800230026245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,7,0.01069760024547577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,7,0.010497599840164185
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,7,0.010548800230026245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,7,0.010489600151777268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,7,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,15,0.012615999579429627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,15,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,15,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,15,0.010977599769830704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,15,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,15,0.010529600083827972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,15,0.010622400045394897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,15,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,15,0.010814400017261505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,15,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,15,0.010611200332641601
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,15,0.010476800054311753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,15,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,15,0.010632000118494033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,31,0.011022400110960007
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,31,0.010876800119876861
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,31,0.012654399871826172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,31,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,31,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,31,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,31,0.010659199953079224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,31,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,31,0.010599999874830245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,31,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,31,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,31,0.010513599961996078
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,31,0.010500799864530563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,31,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,63,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,63,0.012601600587368011
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,63,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,63,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,63,0.010630399733781815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,63,0.010849600285291671
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,63,0.01072160005569458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,63,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,63,0.010675200074911118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,63,0.01061440035700798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,63,0.010520000010728836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,63,0.01061599999666214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,63,0.010585600137710571
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,63,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,127,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,127,0.011243200302124024
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,127,0.010763200372457505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,127,0.010753600299358368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,127,0.010782399773597717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,127,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,127,0.01056319996714592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,127,0.010769599676132202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,127,0.010819199681282043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,127,0.010771200060844421
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,127,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,127,0.010395199805498124
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,127,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,127,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,255,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,255,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,255,0.01109279990196228
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,255,0.010972800105810166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,255,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,255,0.010644800215959548
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,255,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,255,0.010812799632549285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,255,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,255,0.010630399733781815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,255,0.010911999642848969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,255,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,255,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,255,0.010579200088977813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,511,0.01459839940071106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,511,0.01356319934129715
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,511,0.012835200130939483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,511,0.012753599882125854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,511,0.012910400331020356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,511,0.012608000636100769
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,511,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,511,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,511,0.012587200105190276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,511,0.012702399492263794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,511,0.011406400054693223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,511,0.010892800241708755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,511,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,511,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,1023,0.016748799383640288
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,1023,0.014864000678062438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,1023,0.014787200093269347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,1023,0.01470080018043518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,1023,0.013299199938774108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,1023,0.013254399597644805
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,1023,0.01313920021057129
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,1023,0.014723199605941772
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,1023,0.012772800028324127
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,1023,0.012800000607967377
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,1023,0.012833599746227265
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,1023,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,1023,0.01257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,1023,0.012520000338554382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,2047,0.01884640008211136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,2047,0.01680160015821457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,2047,0.014752000570297241
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,2047,0.014683200418949128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,2047,0.014708800613880158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,2047,0.012800000607967377
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,2047,0.012934400141239167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,2047,0.01486240029335022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,2047,0.014758400619029999
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,2047,0.01276479959487915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,2047,0.012702399492263794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,2047,0.012707200646400452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,2047,0.012835200130939483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,2047,0.012753599882125854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,4095,0.02001439929008484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,4095,0.018935999274253844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,4095,0.01868640035390854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,4095,0.01677920073270798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,4095,0.01675039976835251
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,4095,0.016795200109481812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,4095,0.016761599481105803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,4095,0.016913600265979767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,4095,0.0166703999042511
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,4095,0.016729600727558136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,4095,0.014891199767589569
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,4095,0.014716799557209014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,4095,0.01470080018043518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,4095,0.01491519957780838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,8191,0.030851200222969055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,8191,0.020956799387931824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,8191,0.018907199800014495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,8191,0.018931199610233308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,8191,0.018760000169277192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,8191,0.018771199882030486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,8191,0.017948800325393678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,8191,0.022312000393867493
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,8191,0.01865600049495697
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,8191,0.01688639968633652
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,8191,0.016710400581359863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,8191,0.016731199622154237
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,8191,0.016820800304412842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,8191,0.016812799870967864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,16383,0.03115360140800476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,16383,0.03517920076847077
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,16383,0.023523199558258056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,16383,0.023174400627613067
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,16383,0.02287680059671402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,16383,0.02300959974527359
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,16383,0.022940799593925476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,16383,0.029100799560546876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,16383,0.023057599365711213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,16383,0.020718400180339814
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,16383,0.018892799317836762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,16383,0.01903360038995743
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,16383,0.018836799263954162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,16383,0.01892800033092499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,32767,0.0516543984413147
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,32767,0.05339679718017578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,32767,0.04890399873256683
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,32767,0.04776960015296936
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,32767,0.04780159890651703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,32767,0.048728001117706296
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,32767,0.0476639986038208
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,32767,0.033103999495506284
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,32767,0.03121120035648346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,32767,0.0270224004983902
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,32767,0.025179201364517213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,32767,0.026267200708389282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,32767,0.02493920028209686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,32767,0.02510719895362854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,65535,0.07839999794960022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,65535,0.08819040060043334
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,65535,0.07141600251197815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,65535,0.07172960042953491
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,65535,0.07225279808044434
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,65535,0.0725823998451233
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,65535,0.07105439901351929
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,65535,0.04837920069694519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,65535,0.043532800674438474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,65535,0.04186559915542602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,65535,0.04962719976902008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,65535,0.04275679886341095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,65535,0.04351840019226074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,65535,0.04329439997673035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,1,131071,0.13142240047454834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,2,131071,0.15507839918136596
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,4,131071,0.12416000366210937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,8,131071,0.12183040380477905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,16,131071,0.1223520040512085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,32,131071,0.12284320592880249
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,4,1,64,131071,0.12163200378417968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,1,131071,0.070660799741745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,2,131071,0.07894880175590516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,4,131071,0.06593599915504456
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,8,131071,0.06446239948272706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,16,131071,0.0641264021396637
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,32,131071,0.06511520147323609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,4,1,64,131071,0.064164799451828
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,1,0.012647999823093415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,1,0.010609599947929382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,1,0.012622399628162384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,1,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,1,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,1,0.010553599894046783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,1,0.010814400017261505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,1,0.012851199507713318
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,1,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,1,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,1,0.010569600015878677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,1,0.01071999967098236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,1,0.010585600137710571
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,1,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,3,0.01271360069513321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,3,0.012681600451469422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,3,0.010876800119876861
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,3,0.010760000348091126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,3,0.010731200128793717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,3,0.010734400153160096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,3,0.010633599758148194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,3,0.012721599638462066
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,3,0.010704000294208527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,3,0.010745599865913391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,3,0.010598400235176086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,3,0.010552000254392624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,3,0.010782399773597717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,3,0.01058880016207695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,7,0.012639999389648438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,7,0.01098880022764206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,7,0.012652799487113953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,7,0.010731200128793717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,7,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,7,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,7,0.010779199749231338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,7,0.01265919953584671
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,7,0.010662399977445603
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,7,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,7,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,7,0.010956799983978272
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,7,0.010558400303125381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,7,0.01056160032749176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,15,0.012812800705432892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,15,0.012555199861526489
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,15,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,15,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,15,0.0108255997300148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,15,0.0108255997300148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,15,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,15,0.012617599964141846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,15,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,15,0.010735999792814255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,15,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,15,0.010576000064611435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,15,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,15,0.010489600151777268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,31,0.012798400223255157
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,31,0.012577599287033081
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,31,0.011564800143241882
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,31,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,31,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,31,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,31,0.010782399773597717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,31,0.011575999855995178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,31,0.010787200182676315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,31,0.010678400099277497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,31,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,31,0.010486400127410889
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,31,0.010583999752998351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,31,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,63,0.012814399600028992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,63,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,63,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,63,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,63,0.01093600019812584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,63,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,63,0.010523200035095215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,63,0.01133280023932457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,63,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,63,0.010822399705648422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,63,0.01072480008006096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,63,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,63,0.010699199885129929
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,63,0.010649599879980088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,127,0.012785600125789642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,127,0.012649600207805634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,127,0.011617600172758102
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,127,0.012015999853610992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,127,0.010710400342941285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,127,0.010758399963378906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,127,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,127,0.011097600311040878
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,127,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,127,0.010868799686431885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,127,0.010524799674749374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,127,0.010592000186443329
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,127,0.010622400045394897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,127,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,255,0.012627199292182922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,255,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,255,0.01075040027499199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,255,0.011459200084209442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,255,0.010811199992895126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,255,0.01064639985561371
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,255,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,255,0.012676799297332763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,255,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,255,0.010619200021028518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,255,0.010513599961996078
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,255,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,255,0.01075040027499199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,255,0.010675200074911118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,511,0.014718399941921234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,511,0.014708800613880158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,511,0.014745600521564484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,511,0.012905600666999816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,511,0.013068799674510957
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,511,0.012876799702644348
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,511,0.012831999361515046
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,511,0.012758399546146392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,511,0.012545600533485413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,511,0.012593600153923034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,511,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,511,0.012070400267839431
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,511,0.011947199702262878
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,511,0.012625600397586822
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,1023,0.016868799924850464
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,1023,0.016846400499343873
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,1023,0.014753599464893342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,1023,0.014761599898338317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,1023,0.014761599898338317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,1023,0.014718399941921234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,1023,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,1023,0.014902399480342865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,1023,0.014758400619029999
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,1023,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,1023,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,1023,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,1023,0.01281599998474121
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,1023,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,2047,0.020771199464797975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,2047,0.018745599687099455
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,2047,0.016835199296474458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,2047,0.016867199540138246
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,2047,0.016707199811935424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,2047,0.01661120057106018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,2047,0.016710400581359863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,2047,0.014791999757289887
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,2047,0.017343999445438386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,2047,0.014620800316333771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,2047,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,2047,0.012736000120639801
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,2047,0.01271039992570877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,2047,0.01284160017967224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,4095,0.029156801104545594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,4095,0.02089280039072037
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,4095,0.018727999925613404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,4095,0.018769599497318268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,4095,0.018681600689888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,4095,0.01693120002746582
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,4095,0.01685439944267273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,4095,0.02085919976234436
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,4095,0.017028799653053282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,4095,0.016828800737857818
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,4095,0.015024000406265258
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,4095,0.014903999865055084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,4095,0.014774399995803832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,4095,0.014798399806022645
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,8191,0.028092798590660096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,8191,0.032441601157188416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,8191,0.02240000069141388
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,8191,0.020960000157356263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,8191,0.020947200059890748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,8191,0.02152000069618225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,8191,0.020577600598335265
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,8191,0.02503199875354767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,8191,0.02295359969139099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,8191,0.018828800320625304
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,8191,0.018059200048446654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,8191,0.016840000450611115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,8191,0.016782400012016297
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,16383,0.049358400702476504
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,8191,0.016835199296474458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,16383,0.05576159954071045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,16383,0.04580000042915344
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,16383,0.046081599593162534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,16383,0.04579200148582459
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,16383,0.04548319876194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,16383,0.04469760060310364
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,16383,0.02922079861164093
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,16383,0.031227201223373413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,16383,0.024564799666404725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,16383,0.023108799755573273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,16383,0.023057599365711213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,16383,0.02298080027103424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,16383,0.022782400250434875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,32767,0.0760752022266388
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,32767,0.08879520297050476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,32767,0.06889600157737732
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,32767,0.06877599954605103
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,32767,0.06990560293197631
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,32767,0.06917440295219421
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,32767,0.06835839748382569
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,32767,0.04529919922351837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,32767,0.05071200132369995
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,32767,0.0410975992679596
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,32767,0.0393887996673584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,32767,0.039547199010849
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,32767,0.03961600065231323
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,32767,0.03935199975967407
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,65535,0.12788959741592407
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,65535,0.1562351942062378
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,65535,0.12228959798812866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,65535,0.12035200595855713
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,65535,0.1198240041732788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,65535,0.12063519954681397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,65535,0.12011519670486451
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,65535,0.06820319890975952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,65535,0.08076000213623047
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,65535,0.06318719983100891
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,65535,0.061875200271606444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,65535,0.062222397327423094
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,65535,0.06186720132827759
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,65535,0.06144959926605224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,1,131071,0.23165760040283204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,2,131071,0.28955841064453125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,4,131071,0.23666079044342042
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,8,131071,0.2356271982192993
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,16,131071,0.22149920463562012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,32,131071,0.23226399421691896
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,8,1,64,131071,0.2366879940032959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,1,131071,0.11339039802551269
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,2,131071,0.14212640523910522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,4,131071,0.10554399490356445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,8,131071,0.10412479639053344
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,16,131071,0.10343680381774903
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,32,131071,0.1042207956314087
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,8,1,64,131071,0.1049232006072998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,1,0.012582400441169738
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,1,0.01252480000257492
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,1,0.012755200266838074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,1,0.012574400007724761
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,1,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,1,0.01250240057706833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,1,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,1,0.012782399356365205
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,1,0.012759999930858612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,1,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,1,0.01053439974784851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,1,0.010655999928712846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,1,0.010550399869680404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,1,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,3,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,3,0.012608000636100769
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,3,0.012534399330615998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,3,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,3,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,3,0.012110400199890136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,3,0.010887999832630158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,3,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,3,0.012667199969291687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,3,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,3,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,3,0.010657600313425063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,3,0.010795199871063232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,3,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,7,0.012588800489902496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,7,0.01257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,7,0.012652799487113953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,7,0.010856000334024429
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,7,0.01101119965314865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,7,0.010599999874830245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,7,0.010732799768447876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,7,0.012638400495052337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,7,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,7,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,7,0.010435199737548828
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,7,0.010592000186443329
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,7,0.010745599865913391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,7,0.010532800108194351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,15,0.012753599882125854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,15,0.01279360055923462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,15,0.01263359934091568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,15,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,15,0.01098880022764206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,15,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,15,0.011105599999427795
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,15,0.012628799676895142
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,15,0.010820800065994262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,15,0.010788799822330475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,15,0.010660800337791442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,15,0.010790400207042694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,15,0.01064319983124733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,15,0.010496000200510025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,31,0.012756800651550293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,31,0.012729600071907043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,31,0.011644800007343293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,31,0.011190400272607804
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,31,0.010735999792814255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,31,0.012595200538635254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,31,0.010788799822330475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,31,0.012724800407886505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,31,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,31,0.010769599676132202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,31,0.010681600123643876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,31,0.010819199681282043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,31,0.01064319983124733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,31,0.010550399869680404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,63,0.012801599502563477
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,63,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,63,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,63,0.011025600135326385
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,63,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,63,0.011988800019025803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,63,0.010678400099277497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,63,0.012566399574279786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,63,0.012587200105190276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,63,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,63,0.01056160032749176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,63,0.010777600109577179
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,63,0.010623999685049058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,63,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,127,0.012734399735927581
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,127,0.012774400413036346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,127,0.012651200592517852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,127,0.012535999715328216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,127,0.010735999792814255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,127,0.012641599774360657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,127,0.01069599986076355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,127,0.012569600343704223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,127,0.01072480008006096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,127,0.010679999738931656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,127,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,127,0.011342400312423706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,127,0.010664000362157821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,255,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,127,0.010764800012111664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,255,0.01273919939994812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,255,0.012671999633312225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,255,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,255,0.011385600268840789
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,255,0.013036799430847169
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,255,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,255,0.012598399817943574
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,255,0.0110895998775959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,255,0.010791999846696853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,255,0.010531199723482132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,255,0.010545600205659866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,255,0.010734400153160096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,255,0.010619200021028518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,511,0.01483519971370697
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,511,0.014793600142002105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,511,0.014926399290561675
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,511,0.014148800075054169
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,511,0.012851199507713318
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,511,0.012700800597667695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,511,0.012936000525951386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,511,0.014732800424098969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,511,0.01257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,511,0.012580800056457519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,511,0.012572799623012543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,511,0.012563200294971466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,511,0.0107744000852108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,511,0.011224000155925751
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,1023,0.020793600380420683
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,1023,0.01887679994106293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,1023,0.016728000342845918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,1023,0.014822399616241455
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,1023,0.014975999295711518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,1023,0.015254400670528412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,1023,0.01488959938287735
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,1023,0.016812799870967864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,1023,0.015027199685573579
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,1023,0.014620800316333771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,1023,0.012848000228404998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,1023,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,1023,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,1023,0.012863999605178833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,2047,0.028937599062919615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,2047,0.020929600298404693
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,2047,0.018742400407791137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,2047,0.01868479996919632
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,2047,0.017003199458122252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,2047,0.01727840006351471
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,2047,0.017182399332523347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,2047,0.020870399475097657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,2047,0.01674399971961975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,2047,0.014769600331783294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,2047,0.014815999567508698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,2047,0.014753599464893342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,2047,0.014750400185585022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,2047,0.014659200608730317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,4095,0.029078400135040282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,4095,0.029235199093818665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,4095,0.02099200040102005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,4095,0.020934399962425233
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,4095,0.020852799713611602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,4095,0.019495999813079833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,4095,0.018811200559139252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,4095,0.02306240051984787
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,4095,0.020982399582862854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,4095,0.016838400065898894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,4095,0.01672479957342148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,4095,0.016864000260829924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,4095,0.016836799681186676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,4095,0.01675360053777695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,8191,0.05050399899482727
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,8191,0.05208479762077332
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,8191,0.04562560021877289
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,8191,0.04523360133171082
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,8191,0.04474239945411682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,8191,0.04334079921245575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,8191,0.043639999628067014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,8191,0.02906239926815033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,8191,0.02898080050945282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,8191,0.023014399409294128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,8191,0.02099200040102005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,8191,0.021007999777793884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,8191,0.020905600488185884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,8191,0.020897600054740905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,16383,0.07381119728088378
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,16383,0.08947520256042481
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,16383,0.07211999893188477
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,16383,0.07106400132179261
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,16383,0.0723904013633728
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,16383,0.07109439969062806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,16383,0.07047200202941895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,16383,0.0457071989774704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,16383,0.04997439980506897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,16383,0.04049600064754486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,16383,0.0394320011138916
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,16383,0.03941119909286499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,16383,0.03923999965190887
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,16383,0.037920001149177554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,32767,0.12537440061569213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,32767,0.16290559768676757
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,32767,0.12139840126037597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,32767,0.12155519723892212
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,32767,0.12090239524841309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,32767,0.12041280269622803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,32767,0.120033597946167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,32767,0.06829599738121032
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,32767,0.08324480056762695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,32767,0.06227999925613403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,32767,0.06155520081520081
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,32767,0.06128320097923279
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,32767,0.06082239747047424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,32767,0.06098560094833374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,65535,0.22948799133300782
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,65535,0.31004319190979
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,65535,0.2193120002746582
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,65535,0.22278239727020263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,65535,0.22425920963287355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,65535,0.21793599128723146
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,65535,0.21711199283599852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,65535,0.11386079788208008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,65535,0.1517135977745056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,65535,0.10525599718093873
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,65535,0.1042464017868042
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,65535,0.10510400533676148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,65535,0.10487040281295776
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,65535,0.10321279764175414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,1,131071,0.4349952220916748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,2,131071,0.6022784233093261
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,4,131071,0.42403359413146974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,8,131071,0.4191152095794678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,16,131071,0.41883039474487305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,32,131071,0.42034077644348145
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,16,1,64,131071,0.42173118591308595
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,1,131071,0.20670878887176514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,2,131071,0.2839184045791626
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,4,131071,0.1914736032485962
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,8,131071,0.19158240556716918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,16,131071,0.19088159799575805
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,32,131071,0.19041759967803956
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,16,1,64,131071,0.19069279432296754
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,1,0.014608000218868256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,1,0.012936000525951386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,1,0.012824000418186187
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,1,0.012569600343704223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,1,0.012748800218105316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,1,0.012718400359153748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,1,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,1,0.012756800651550293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,1,0.013107199966907502
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,1,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,1,0.010804799944162368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,1,0.010807999968528747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,1,0.012566399574279786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,1,0.010945600271224976
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,3,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,3,0.012915199995040894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,3,0.012598399817943574
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,3,0.012601600587368011
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,3,0.012699200212955475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,3,0.012515200674533844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,3,0.0126351997256279
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,3,0.01297920048236847
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,3,0.012758399546146392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,3,0.01257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,3,0.010715200006961823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,3,0.012593600153923034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,3,0.010953599959611893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,3,0.011452800035476685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,7,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,7,0.012889599800109864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,7,0.012732799351215362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,7,0.012656000256538392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,7,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,7,0.012544000148773193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,7,0.012585599720478059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,7,0.012801599502563477
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,7,0.012577599287033081
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,7,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,7,0.010691200196743012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,7,0.01075040027499199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,7,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,7,0.010681600123643876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,15,0.014678399264812469
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,15,0.012822400033473968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,15,0.012664000689983367
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,15,0.012510399520397186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,15,0.012600000202655792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,15,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,15,0.012617599964141846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,15,0.012865599989891053
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,15,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,15,0.010966400057077408
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,15,0.010891199856996537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,15,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,15,0.010676799714565277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,15,0.010758399963378906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,31,0.014732800424098969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,31,0.01273919939994812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,31,0.012768000364303589
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,31,0.012559999525547028
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,31,0.012539200484752655
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,31,0.012479999661445617
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,31,0.012707200646400452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,31,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,31,0.012606400251388549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,31,0.011273600161075592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,31,0.010836800187826156
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,31,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,31,0.010824000090360641
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,31,0.010787200182676315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,63,0.014796799421310425
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,63,0.013073599338531494
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,63,0.012828800082206725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,63,0.012785600125789642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,63,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,63,0.012726399302482604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,63,0.01255200058221817
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,63,0.012848000228404998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,63,0.012654399871826172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,63,0.012582400441169738
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,63,0.010843200236558914
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,63,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,63,0.011521600186824799
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,63,0.010657600313425063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,127,0.014803199470043183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,127,0.012750400602817536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,127,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,127,0.012828800082206725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,127,0.013075199723243714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,127,0.012614400684833526
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,127,0.012860800325870513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,127,0.012827199697494508
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,127,0.012641599774360657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,127,0.010660800337791442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,127,0.010838399827480315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,127,0.01077279970049858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,127,0.010812799632549285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,127,0.010662399977445603
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,255,0.01467359960079193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,255,0.01284479945898056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,255,0.012742400169372559
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,255,0.012788799405097962
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,255,0.012649600207805634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,255,0.01265760064125061
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,255,0.01265919953584671
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,255,0.012670400738716125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,255,0.012726399302482604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,255,0.011326400190591812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,255,0.010807999968528747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,255,0.01072160005569458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,255,0.010681600123643876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,255,0.010713600367307664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,511,0.018833599984645844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,511,0.016836799681186676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,511,0.015006400644779205
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,511,0.014711999893188476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,511,0.014902399480342865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,511,0.014747199416160584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,511,0.014748799800872802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,511,0.014742399752140044
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,511,0.014668799936771393
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,511,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,511,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,511,0.012625600397586822
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,511,0.012510399520397186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,511,0.012755200266838074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,1023,0.025225600600242613
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,1023,0.020937600731849672
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,1023,0.01676799952983856
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,1023,0.01685439944267273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,1023,0.016835199296474458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,1023,0.016774399578571318
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,1023,0.016857600212097167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,1023,0.018900799751281738
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,1023,0.016804799437522888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,1023,0.014788800477981567
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,1023,0.014620800316333771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,1023,0.014684799313545226
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,1023,0.014828799664974213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,1023,0.014630399644374847
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,2047,0.027187201380729675
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,2047,0.03107360005378723
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,2047,0.02096959948539734
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,2047,0.020476800203323365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,2047,0.020838400721549986
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,2047,0.018913599848747253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,2047,0.020777599513530733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,2047,0.020904000103473663
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,2047,0.02096800059080124
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,2047,0.01674560010433197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,2047,0.01679839938879013
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,2047,0.014868800342082978
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,2047,0.014918400347232819
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,2047,0.014787200093269347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,4095,0.0496832013130188
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,4095,0.05441120266914368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,4095,0.04338879883289337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,4095,0.04213440120220184
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,4095,0.042475199699401854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,4095,0.04147520065307617
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,4095,0.04171839952468872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,4095,0.02698720097541809
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,4095,0.029102399945259094
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,4095,0.020927999913692475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,4095,0.018956799805164338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,4095,0.018990400433540344
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,4095,0.019091199338436126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,4095,0.018668800592422485
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,8191,0.07554240226745605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,8191,0.0910319983959198
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,8191,0.07041760087013245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,8191,0.06940640211105346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,8191,0.06809759736061097
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,8191,0.06875360012054443
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,8191,0.06845759749412536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,8191,0.045440000295639035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,8191,0.05116320252418518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,8191,0.03776159882545471
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,8191,0.037268799543380735
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,8191,0.03583360016345978
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,8191,0.03678080141544342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,8191,0.035867199301719666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,16383,0.12885760068893432
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,16383,0.16468000411987305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,16383,0.12130880355834961
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,16383,0.11937919855117798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,16383,0.11947519779205322
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,16383,0.11943039894104004
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,16383,0.1200976014137268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,16383,0.07050880193710327
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,16383,0.08489120006561279
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,16383,0.06114879846572876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,16383,0.05984479784965515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,16383,0.05969280004501343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,16383,0.06021760106086731
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,16383,0.0594543993473053
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,32767,0.233076810836792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,32767,0.3103631973266602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,32767,0.22665119171142578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,32767,0.22453761100769043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,32767,0.22548320293426513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,32767,0.2258080005645752
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,32767,0.22620160579681398
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,32767,0.11904640197753906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,32767,0.15245280265808106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,32767,0.10672960281372071
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,32767,0.10537760257720948
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,32767,0.10499039888381959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,32767,0.10497759580612183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,32767,0.10466079711914063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,1,65535,0.4389664173126221
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,2,65535,0.6023808002471924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,4,65535,0.4320847988128662
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,8,65535,0.4291264057159424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,16,65535,0.4305312156677246
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,32,65535,0.43094401359558104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,32,1,64,65535,0.4340688228607178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,1,65535,0.21554241180419922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,2,65535,0.2861936092376709
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,4,65535,0.19561920166015626
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,8,65535,0.19396959543228148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,16,65535,0.19432640075683594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,32,65535,0.19359200000762938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,32,1,64,65535,0.1944208025932312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,1,0.014662399888038635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,1,0.012694400548934937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,1,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,1,0.012804800271987915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,1,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,1,0.012601600587368011
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,1,0.012694400548934937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,1,0.01268479973077774
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,1,0.013867199420928955
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,1,0.012680000066757202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,1,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,1,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,1,0.012825599312782288
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,1,0.010713600367307664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,3,0.01480959951877594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,3,0.0146479994058609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,3,0.01289760023355484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,3,0.012617599964141846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,3,0.012915199995040894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,3,0.012612800300121307
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,3,0.012886400520801543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,3,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,3,0.01284639984369278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,3,0.012636800110340119
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,3,0.010815999656915664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,3,0.010756800323724747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,3,0.010815999656915664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,3,0.010689599812030793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,7,0.014679999649524688
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,7,0.01467519998550415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,7,0.012838399410247803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,7,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,7,0.012950399518013
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,7,0.012800000607967377
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,7,0.012814399600028992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,7,0.01284479945898056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,7,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,7,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,7,0.010815999656915664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,7,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,7,0.010844799876213073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,7,0.01085119992494583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,15,0.014611199498176575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,15,0.014742399752140044
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,15,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,15,0.01250240057706833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,15,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,15,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,15,0.012849600613117218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,15,0.012929600477218629
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,15,0.01266240030527115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,15,0.012694400548934937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,15,0.012608000636100769
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,15,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,15,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,15,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,31,0.014764800667762756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,31,0.014747199416160584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,31,0.013604800403118133
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,31,0.012777599692344665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,31,0.012870399653911591
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,31,0.012812800705432892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,31,0.012718400359153748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,31,0.01282079964876175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,31,0.012807999551296235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,31,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,31,0.010817600041627884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,31,0.010859200358390808
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,31,0.010856000334024429
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,31,0.010763200372457505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,63,0.014577600359916686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,63,0.014864000678062438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,63,0.012915199995040894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,63,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,63,0.012852799892425538
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,63,0.012870399653911591
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,63,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,63,0.012995199859142303
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,63,0.01279519945383072
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,63,0.012652799487113953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,63,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,63,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,63,0.010735999792814255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,63,0.010820800065994262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,127,0.014711999893188476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,127,0.014627200365066529
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,127,0.012848000228404998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,127,0.0127920001745224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,127,0.01292479932308197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,127,0.012800000607967377
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,127,0.012948800623416901
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,127,0.01268479973077774
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,127,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,127,0.012615999579429627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,127,0.01080000028014183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,127,0.010833600163459777
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,127,0.010756800323724747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,127,0.010779199749231338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,255,0.014947199821472168
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,255,0.014883199334144592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,255,0.012449599802494049
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,255,0.012788799405097962
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,255,0.012812800705432892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,255,0.01282079964876175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,255,0.012736000120639801
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,255,0.014628799259662628
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,255,0.012775999307632447
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,255,0.012641599774360657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,255,0.010694400221109391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,255,0.010777600109577179
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,255,0.010875199735164643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,255,0.011009600013494492
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,511,0.02300640046596527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,511,0.01892800033092499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,511,0.016791999340057373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,511,0.01663520038127899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,511,0.014803199470043183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,511,0.015846399962902068
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,511,0.016835199296474458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,511,0.021118399500846863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,511,0.014662399888038635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,511,0.01271360069513321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,511,0.012848000228404998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,511,0.012652799487113953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,511,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,511,0.012744000554084778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,1023,0.029142400622367857
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,1023,0.029236799478530882
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,1023,0.02072480022907257
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,1023,0.01887200027704239
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,1023,0.018795199692249298
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,1023,0.018801599740982056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,1023,0.01876160055398941
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,1023,0.02311519980430603
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,1023,0.018855999410152435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,1023,0.016888000071048737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,1023,0.014788800477981567
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,1023,0.01478240042924881
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,1023,0.015022400021553039
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,1023,0.014752000570297241
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,2047,0.05211039781570435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,2047,0.052160000801086424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,2047,0.0447952002286911
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,2047,0.04351359903812409
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,2047,0.041700801253318785
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,2047,0.04145120084285736
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,2047,0.04294080138206482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,2047,0.029023998975753786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,2047,0.02614560127258301
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,2047,0.020900799334049223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,2047,0.01923999935388565
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,2047,0.018838399648666383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,2047,0.018804800510406495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,2047,0.018918399512767792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,4095,0.0791647970676422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,4095,0.07169280052185059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,4095,0.0886080026626587
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,4095,0.07112640142440796
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,4095,0.0699679970741272
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,4095,0.07006080150604248
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,4095,0.06899679899215698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,4095,0.04812000095844269
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,4095,0.049721598625183105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,4095,0.0391728013753891
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,4095,0.03746559917926788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,4095,0.03692159950733185
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,4095,0.03748640120029449
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,4095,0.037363201379776
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,8191,0.13262879848480225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,8191,0.16035840511322022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,8191,0.12295680046081543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,8191,0.11976159811019897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,8191,0.12133920192718506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,8191,0.11943999528884888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,8191,0.11987680196762085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,8191,0.07238720059394836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,8191,0.08282080292701721
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,8191,0.06197919845581055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,8191,0.061289602518081666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,8191,0.059601598978042604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,8191,0.059703999757766725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,8191,0.05922080278396606
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,16383,0.23733279705047608
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,16383,0.304582405090332
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,16383,0.22940480709075928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,16383,0.22451679706573485
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,16383,0.22524480819702147
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,16383,0.22378880977630616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,16383,0.2271359920501709
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,16383,0.14973440170288085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,16383,0.10684959888458252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,16383,0.10520319938659668
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,16383,0.12281759977340698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,16383,0.10528800487518311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,16383,0.10499360561370849
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,16383,0.10468480587005616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,1,32767,0.44060320854187013
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,2,32767,0.5925871849060058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,4,32767,0.43584480285644533
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,8,32767,0.4312751770019531
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,16,32767,0.43031997680664064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,32,32767,0.4282815933227539
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,64,1,64,32767,0.4337615966796875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,1,32767,0.21742560863494872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,2,32767,0.28590400218963624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,4,32767,0.19697279930114747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,16,32767,0.19377280473709108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,8,32767,0.19453439712524415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,32,32767,0.19430240392684936
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,64,1,64,32767,0.19377599954605101
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,1,0.020742399990558623
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,1,0.01483680009841919
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,1,0.016683200001716615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,1,0.01271360069513321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,1,0.014767999947071075
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,1,0.014683200418949128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,1,0.01313599944114685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,1,0.014777599275112152
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,1,0.0176256000995636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,1,0.01284639984369278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,1,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,1,0.012636800110340119
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,1,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,1,0.01263200044631958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,3,0.016761599481105803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,3,0.020742399990558623
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,3,0.014787200093269347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,3,0.014644800126552582
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,3,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,3,0.014187200367450714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,3,0.014608000218868256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,3,0.016780799627304076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,3,0.01847199946641922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,3,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,3,0.012775999307632447
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,3,0.012548799812793731
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,3,0.012622399628162384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,3,0.010920000076293946
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,7,0.016785599291324615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,7,0.02086720019578934
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,7,0.01467519998550415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,7,0.013652800023555756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,7,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,7,0.014696000516414643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,7,0.014484800398349762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,7,0.01483680009841919
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,7,0.017531199753284453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,7,0.012779200077056884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,7,0.012667199969291687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,7,0.012567999958992004
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,7,0.012585599720478059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,7,0.012539200484752655
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,15,0.016711999475955964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,15,0.020824000239372253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,15,0.0147599995136261
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,15,0.014616000652313232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,15,0.015062400698661804
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,15,0.014678399264812469
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,15,0.01494079977273941
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,15,0.016760000586509706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,15,0.017334400117397307
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,15,0.012761600315570831
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,15,0.0126351997256279
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,15,0.012587200105190276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,15,0.012656000256538392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,15,0.012577599287033081
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,31,0.01690399944782257
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,31,0.02088959962129593
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,31,0.014815999567508698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,31,0.014232000708580017
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,31,0.01465120017528534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,31,0.014681600034236908
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,31,0.014564800262451171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,31,0.016392000019550323
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,31,0.018294399976730345
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,31,0.012750400602817536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,31,0.012727999687194824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,31,0.012604799866676331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,31,0.01263200044631958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,31,0.012664000689983367
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,63,0.016784000396728515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,63,0.020844799280166627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,63,0.014886400103569031
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,63,0.014587199687957764
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,63,0.01462559998035431
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,63,0.014668799936771393
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,63,0.01459999978542328
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,63,0.016755199432373045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,63,0.017132799327373504
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,63,0.012895999848842621
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,63,0.012604799866676331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,63,0.012697599828243256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,63,0.01268479973077774
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,63,0.01263359934091568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,127,0.016814400255680085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,127,0.02093919962644577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,127,0.014752000570297241
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,127,0.014692799746990204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,127,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,127,0.014731200039386749
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,127,0.013249599933624267
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,127,0.016785599291324615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,127,0.017609600722789765
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,127,0.012886400520801543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,127,0.012596799433231354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,127,0.012731200456619263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,127,0.012580800056457519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,127,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,255,0.018830400705337525
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,255,0.020838400721549986
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,255,0.01472959965467453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,255,0.014689600467681885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,255,0.014720000326633453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,255,0.014635199308395385
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,255,0.014654399454593658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,255,0.01681919991970062
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,255,0.017803199589252472
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,255,0.012724800407886505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,255,0.012615999579429627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,255,0.012430399656295776
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,255,0.012608000636100769
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,255,0.012583999335765839
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,511,0.025979200005531312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,511,0.03100320100784302
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,511,0.01894560009241104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,511,0.018727999925613404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,511,0.018435199558734894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,511,0.018265600502490997
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,511,0.017015999555587767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,511,0.018915200233459474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,511,0.02093600034713745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,511,0.014880000054836274
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,511,0.013913600146770478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,511,0.014596800506114959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,511,0.014758400619029999
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,511,0.014595200121402741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,1023,0.04828960001468659
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,1023,0.05378720164299011
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,1023,0.041950398683547975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,1023,0.041280001401901245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,1023,0.04036000072956085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,1023,0.04112800061702728
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,1023,0.03999359905719757
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,1023,0.02558560073375702
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,1023,0.02985920011997223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,1023,0.019020800292491914
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,1023,0.01879359930753708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,1023,0.01732639968395233
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,1023,0.01713760048151016
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,1023,0.016891199350357055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,2047,0.0753216028213501
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,2047,0.09315199851989746
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,2047,0.06914079785346985
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,2047,0.06652960181236267
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,2047,0.06711999773979187
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,2047,0.06778560280799865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,2047,0.06678239703178405
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,2047,0.04551520049571991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,2047,0.051939201354980466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,2047,0.037339198589324954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,2047,0.0365231990814209
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,2047,0.03630400002002716
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,2047,0.03573600053787231
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,2047,0.03586559891700745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,4095,0.12623519897460939
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,4095,0.16706399917602538
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,4095,0.11930240392684936
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,4095,0.11838719844818116
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,4095,0.1178879976272583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,4095,0.11707040071487426
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,4095,0.1182528018951416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,4095,0.07030720114707947
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,4095,0.08617920279502869
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,4095,0.06000319719314575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,4095,0.05956320166587829
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,4095,0.05956159830093384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,4095,0.05803999900817871
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,4095,0.0580672025680542
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,8191,0.23473920822143554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,8191,0.31274240016937255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,8191,0.22377281188964843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,8191,0.22225279808044435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,8191,0.22398879528045654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,8191,0.2217103958129883
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,8191,0.22226080894470215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,8191,0.11891839504241944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,8191,0.15343999862670898
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,8191,0.10649600028991699
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,8191,0.10470080375671387
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,8191,0.10402560234069824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,8191,0.10485119819641113
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,8191,0.10402400493621826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,1,16383,0.44185600280761717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,2,16383,0.602953577041626
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,4,16383,0.4274400234222412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,8,16383,0.4253727912902832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,16,16383,0.42852158546447755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,32,16383,0.42829442024230957
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,128,1,64,16383,0.42455039024353025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,1,16383,0.21358559131622315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,2,16383,0.29010560512542727
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,4,16383,0.19403040409088135
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,8,16383,0.19372639656066895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,16,16383,0.19265120029449462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,32,16383,0.19225599765777587
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,128,1,64,16383,0.19226239919662474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,1,0.024987199902534486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,1,0.031070399284362792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,1,0.02083200067281723
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,1,0.018905599415302277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,1,0.018904000520706177
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,1,0.019083200395107268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,1,0.018984000384807586
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,1,0.023078399896621703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,1,0.0181536003947258
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,1,0.02688319981098175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,1,0.01669919937849045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,1,0.016737599670886994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,1,0.016340799629688263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,1,0.016760000586509706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,3,0.024879999458789825
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,3,0.031179198622703554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,3,0.020776000618934632
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,3,0.01903039962053299
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,3,0.019492800533771514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,3,0.018958400189876556
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,3,0.018828800320625304
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,3,0.023107199370861052
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,3,0.025094398856163026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,3,0.018775999546051025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,3,0.016689600050449373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,3,0.016927999258041383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,3,0.016475200653076172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,3,0.01642879992723465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,7,0.024902400374412537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,7,0.02075680047273636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,7,0.031159999966621398
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,7,0.019115200638771056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,7,0.01913760006427765
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,7,0.01918399930000305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,7,0.019097599387168884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,7,0.022944000363349915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,7,0.025142401456832886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,7,0.018580800294876097
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,7,0.01677920073270798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,7,0.016524800658226015
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,7,0.0165583997964859
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,7,0.016811199486255646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,15,0.024935999512672426
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,15,0.031227201223373413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,15,0.020878399908542632
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,15,0.020510399341583253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,15,0.01919199973344803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,15,0.019849599897861482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,15,0.01937440037727356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,15,0.023056000471115112
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,15,0.026910400390625
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,15,0.018756799399852753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,15,0.016624000668525696
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,15,0.016616000235080718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,15,0.016795200109481812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,15,0.016785599291324615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,31,0.02496960014104843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,31,0.0312032014131546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,31,0.02106720060110092
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,31,0.019735999405384064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,31,0.019200000166893005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,31,0.01969120055437088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,31,0.01926079988479614
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,31,0.023134399950504304
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,31,0.025227200984954835
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,31,0.018727999925613404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,31,0.01690080016851425
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,31,0.016760000586509706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,31,0.01636800020933151
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,31,0.016715200245380403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,63,0.024966399371623992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,63,0.03136000037193298
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,63,0.020880000293254854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,63,0.019862399995326997
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,63,0.020121599733829498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,63,0.019278399646282196
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,63,0.01913760006427765
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,63,0.022993600368499754
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,63,0.027348798513412476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,63,0.01879040002822876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,63,0.01671839952468872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,63,0.01671359986066818
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,63,0.016817599534988403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,63,0.01650400012731552
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,127,0.025076800584793092
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,127,0.03118399977684021
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,127,0.02091200053691864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,127,0.0190528005361557
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,127,0.019270400702953338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,127,0.019276799261569978
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,127,0.01982239931821823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,127,0.022991999983787537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,127,0.025059199333190917
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,127,0.018731200695037843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,127,0.01669120043516159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,127,0.01655679941177368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,127,0.016808000206947327
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,127,0.01671359986066818
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,255,0.03725599944591522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,255,0.03483999967575073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,255,0.021169599890708924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,255,0.02048960030078888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,255,0.019383999705314636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,255,0.019817599654197694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,255,0.020502400398254395
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,255,0.027105599641799927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,255,0.025198400020599365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,255,0.01849440038204193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,255,0.016940799355506898
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,255,0.016704000532627106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,255,0.016684800386428833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,255,0.016223999857902526
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,511,0.053648000955581664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,511,0.058064001798629764
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,511,0.044310399889945985
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,511,0.0421887993812561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,511,0.04162079989910126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,511,0.04143039882183075
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,511,0.04114879965782166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,511,0.03750079870223999
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,511,0.036743998527526855
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,511,0.022737599909305573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,511,0.02062080055475235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,511,0.01969760060310364
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,511,0.020771199464797975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,511,0.020319999754428865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,1023,0.08211680054664612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,1023,0.09567520022392273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,1023,0.07092959880828857
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,1023,0.06830559968948365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,1023,0.06782240271568299
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,1023,0.06905279755592346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,1023,0.06737120151519775
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,1023,0.05490080118179321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,1023,0.04014880061149597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,1023,0.05175840258598328
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,1023,0.03757599890232086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,1023,0.03707039952278137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,1023,0.03700959980487824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,1023,0.03739520013332367
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,2047,0.1379680037498474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,2047,0.17047359943389892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,2047,0.12380000352859497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,2047,0.12136640548706054
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,2047,0.12105760574340821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,2047,0.12026079893112182
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,2047,0.12042080163955689
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,2047,0.07879040241241456
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,2047,0.0902176022529602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,2047,0.06381120085716248
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,2047,0.06119840145111084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,2047,0.06066399812698364
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,2047,0.060185599327087405
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,2047,0.059671998023986816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,4095,0.2422719955444336
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,4095,0.31761600971221926
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,4095,0.23054399490356445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,4095,0.2249903917312622
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,4095,0.22368960380554198
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,4095,0.22734079360961915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,4095,0.22541120052337646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,4095,0.12782080173492433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,4095,0.1596287965774536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,4095,0.10914880037307739
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,4095,0.10700800418853759
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,4095,0.10654079914093018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,4095,0.10548959970474243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,4095,0.10505280494689942
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,1,8191,0.4522863864898682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,2,8191,0.6080080032348633
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,4,8191,0.43323841094970705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,8,8191,0.43030400276184083
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,16,8191,0.4291264057159424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,32,8191,0.43324480056762693
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,256,1,64,8191,0.4291664123535156
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,1,8191,0.2268671989440918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,2,8191,0.2932735919952393
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,4,8191,0.19933760166168213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,8,8191,0.19712159633636475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,16,8191,0.19697760343551635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,32,8191,0.1961967945098877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,256,1,64,8191,0.1950559973716736
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,1,0.043289598822593686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,1,0.049644801020622256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,1,0.030937600135803222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,1,0.02918879985809326
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,1,0.029212799668312073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,1,0.029051199555397034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,1,0.028977599740028382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,1,0.03742400109767914
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,1,0.042315199971199036
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,1,0.026966398954391478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,1,0.022940799593925476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,1,0.022993600368499754
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,1,0.02298559993505478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,3,0.04237279891967773
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,3,0.049558401107788086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,1,0.022963200509548188
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,3,0.03127360045909881
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,3,0.029084798693656922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,3,0.029131200909614564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,3,0.029041600227355958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,3,0.029155200719833373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,3,0.03741439878940582
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,3,0.04176000058650971
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,3,0.02699359953403473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,3,0.024817599356174468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,3,0.022969600558280946
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,3,0.023155200481414794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,3,0.02287680059671402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,7,0.04333280026912689
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,7,0.04973919987678528
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,7,0.03123520016670227
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,7,0.02916960120201111
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,7,0.02924000024795532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,7,0.029153600335121155
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,7,0.02918879985809326
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,7,0.0381520003080368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,7,0.04273920059204102
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,7,0.027020800113677978
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,7,0.02359360009431839
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,7,0.02298080027103424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,7,0.022942399978637694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,7,0.022974400222301482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,15,0.04336000084877014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,15,0.05068640112876892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,15,0.03196640014648437
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,15,0.02914400100708008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,15,0.02928000092506409
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,15,0.029131200909614564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,15,0.02916960120201111
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,15,0.037329599261283875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,15,0.04317600131034851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,15,0.026972800493240356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,15,0.024846400320529937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,15,0.02312159985303879
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,15,0.02297919988632202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,15,0.022812800109386445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,31,0.043803200125694275
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,31,0.05022720098495483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,31,0.031481599807739256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,31,0.02921760082244873
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,31,0.02928000092506409
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,31,0.029262399673461913
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,31,0.029131200909614564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,31,0.03923999965190887
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,31,0.04297119975090027
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,31,0.027108800411224366
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,31,0.023148800432682037
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,31,0.023003199696540834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,31,0.023185600340366364
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,31,0.02295520007610321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,63,0.04547840058803558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,63,0.05179200172424316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,63,0.03153280019760132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,63,0.029068800806999206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,63,0.029158401489257812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,63,0.02908160090446472
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,63,0.029177600145339967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,63,0.03930400013923645
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,63,0.027225598692893982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,63,0.024795199930667877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,63,0.042840000987052915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,63,0.02289759963750839
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,63,0.02290080040693283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,63,0.022950400412082673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,127,0.049553599953651425
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,127,0.054743999242782594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,127,0.033395200967788696
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,127,0.029635199904441835
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,127,0.028935998678207397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,127,0.02938399910926819
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,127,0.02913280129432678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,127,0.04158239960670471
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,127,0.04318400025367737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,127,0.026966398954391478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,127,0.022950400412082673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,127,0.023127999901771546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,127,0.022950400412082673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,127,0.02295680046081543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,255,0.06532480120658875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,255,0.06026560068130493
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,255,0.04743039906024933
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,255,0.04350399971008301
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,255,0.0433023989200592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,255,0.04195199906826019
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,255,0.041993600130081174
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,255,0.0495855987071991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,255,0.04591200053691864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,255,0.027163198590278624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,255,0.024059200286865236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,255,0.02292640060186386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,255,0.023076799511909486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,255,0.022870400547981264
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,511,0.09610080122947692
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,511,0.09564639925956726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,511,0.07800319790840149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,511,0.07171040177345275
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,511,0.07207520008087158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,511,0.0703279972076416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,511,0.0713487982749939
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,511,0.06376799941062927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,511,0.05996320247650146
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,511,0.0435696005821228
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,511,0.04144960045814514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,511,0.039299198985099794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,511,0.03946399986743927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,511,0.038761600852012634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,1023,0.1488976001739502
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,1023,0.16344799995422363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,1023,0.1285423994064331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,1023,0.12436319589614868
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,1023,0.1240656018257141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,1023,0.12207200527191162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,1023,0.1226304054260254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,1023,0.08901119828224183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,1023,0.09210079908370972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,1023,0.06800000071525573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,1023,0.06400799751281738
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,1023,0.06269760131835937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,1023,0.06208320260047913
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,1023,0.06154879927635193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,2047,0.26479520797729494
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,2047,0.2990560054779053
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,2047,0.23551199436187745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,2047,0.22992639541625975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,2047,0.23199200630187988
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,2047,0.2318336009979248
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,2047,0.22745919227600098
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,2047,0.1404096007347107
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,2047,0.15443999767303468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,2047,0.11407519578933716
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,2047,0.11072640419006348
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,2047,0.10873119831085205
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,2047,0.10858559608459473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,2047,0.10752639770507813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,1,4095,0.4811039924621582
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,2,4095,0.5577487945556641
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,4,4095,0.44513440132141113
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,8,4095,0.44024319648742677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,16,4095,0.4386928081512451
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,32,4095,0.4388879776000977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,512,1,64,4095,0.4352287769317627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,1,4095,0.24136641025543212
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,2,4095,0.27887840270996095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,4,4095,0.2076416015625
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,8,4095,0.2021343946456909
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,16,4095,0.20062079429626464
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,512,1,64,4095,0.19899200201034545
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,32,4095,0.19992799758911134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,1,0.07439200282096863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,1,0.09455040097236633
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,1,0.0514415979385376
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,1,0.04763039946556091
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,1,0.045767998695373534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,1,0.04579519927501678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,1,0.0455951988697052
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,1,0.06889119744300842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,1,0.07634400129318238
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,1,0.04350239932537079
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,1,0.03715679943561554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,1,0.036857599020004274
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,1,0.035278400778770445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,1,0.03516640067100525
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,3,0.0743664026260376
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,3,0.09493759870529175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,3,0.05167040228843689
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,3,0.04548319876194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,3,0.0474480003118515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,3,0.04700320065021515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,3,0.04582079946994781
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,3,0.07609120011329651
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,3,0.06857759952545166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,3,0.04351840019226074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,3,0.03735679984092712
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,3,0.035369598865509035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,3,0.035252800583839415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,3,0.03528960049152374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,7,0.0742143988609314
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,7,0.05163679718971252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,7,0.0474047988653183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,7,0.09460480213165283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,7,0.045691201090812684
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,7,0.04608800113201141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,7,0.047131198644638064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,7,0.06859359741210938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,7,0.07625759840011596
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,7,0.043459200859069826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,7,0.03747999966144562
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,7,0.03542239964008331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,7,0.03534719944000244
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,7,0.03526560068130493
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,15,0.07415199875831605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,15,0.09481920003890991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,15,0.05276319980621338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,15,0.04751999974250794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,15,0.0474368005990982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,15,0.04747200012207031
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,15,0.04607360064983368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,15,0.06984959840774536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,15,0.07632480263710022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,15,0.04420480132102966
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,15,0.037257599830627444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,15,0.03708640038967133
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,15,0.03534240126609802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,15,0.035308799147605895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,31,0.07493919730186463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,31,0.09513760209083558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,31,0.053744000196456906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,31,0.04735519886016846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,31,0.047470399737358095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,31,0.04744159877300262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,31,0.047270399332046506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,31,0.07008159756660462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,31,0.07810720205307006
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,31,0.04395680129528046
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,31,0.03733119964599609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,31,0.03531999886035919
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,31,0.03521760106086731
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,31,0.035288000106811525
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,63,0.07693279981613159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,63,0.09454240202903748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,63,0.05605760216712952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,63,0.04965600073337555
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,63,0.0477183997631073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,63,0.04747200012207031
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,63,0.04640640020370483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,63,0.07006719708442688
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,63,0.0783568024635315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,63,0.044860801100730895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,63,0.03716799914836884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,63,0.03711200058460236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,63,0.03520799875259399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,63,0.03526400029659271
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,127,0.08664159774780274
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,127,0.09718559980392456
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,127,0.06418560147285461
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,127,0.05655999779701233
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,127,0.055129599571228025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,127,0.05379840135574341
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,127,0.05351200103759766
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,127,0.07015200257301331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,127,0.07832000255584717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,127,0.04744159877300262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,127,0.03743039965629578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,127,0.03584319949150085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,127,0.03534719944000244
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,127,0.03526400029659271
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,255,0.11169279813766479
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,255,0.10665119886398315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,255,0.07743200063705444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,255,0.07218719720840454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,255,0.0702351987361908
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,255,0.06867520213127136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,255,0.06823999881744384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,255,0.08338080048561096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,255,0.07710080146789551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,255,0.05106080174446106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,255,0.045484799146652224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,255,0.04285280108451843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,255,0.04160319864749908
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,255,0.04115839898586273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,511,0.1628208041191101
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,511,0.17765120267868043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,511,0.13348480463027954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,511,0.12587679624557496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,511,0.1210752010345459
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,511,0.12154719829559327
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,511,0.12340960502624512
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,511,0.10488320589065551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,511,0.10575040578842163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,511,0.07203199863433837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,511,0.06633920073509217
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,511,0.06379839777946472
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,511,0.061724799871444705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,511,0.062428802251815796
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,1023,0.26541121006011964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,1023,0.3101615905761719
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,1023,0.2262255907058716
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,1023,0.2196880102157593
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,1023,0.21788160800933837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,1023,0.21452479362487792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,1023,0.21359519958496093
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,1023,0.15468159914016724
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,1023,0.16806399822235107
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,1023,0.11659679412841797
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,1023,0.10950080156326295
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,1023,0.10699520111083985
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,1023,0.10565439462661744
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,1023,0.10485279560089111
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,1,2047,0.47847681045532225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,4,2047,0.43065438270568845
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,2,2047,0.5798272132873535
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,8,2047,0.42363681793212893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,16,2047,0.41972641944885253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,32,2047,0.4173696041107178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,1,1024,1,64,2047,0.4144768238067627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,1,2047,0.2502351999282837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,2,2047,0.2924079895019531
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,4,2047,0.20540800094604492
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,8,2047,0.19708479642868043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,16,2047,0.1933743953704834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,32,2047,0.19090720415115356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,1,0.012563200294971466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,1,1024,1,64,2047,0.18973920345306397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,1,0.010702399909496308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,1,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,1,0.0107744000852108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,1,0.010729599744081497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,1,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,1,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,1,0.010892800241708755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,1,0.010649599879980088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,1,0.010548800230026245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,1,0.010665600001811982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,1,0.010598400235176086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,1,0.010451199859380722
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,1,0.010548800230026245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,3,0.011671999841928482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,3,0.010779199749231338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,3,0.011558400094509124
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,3,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,3,0.010779199749231338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,3,0.010627199709415436
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,3,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,3,0.010633599758148194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,3,0.01085280030965805
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,3,0.010632000118494033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,3,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,3,0.009431999921798707
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,3,0.01058880016207695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,3,0.010547199845314026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,7,0.012596799433231354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,7,0.010868799686431885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,7,0.012670400738716125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,7,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,7,0.011308799684047698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,7,0.010540799796581268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,7,0.010923200100660325
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,7,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,7,0.0108815997838974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,7,0.010529600083827972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,7,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,7,0.010644800215959548
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,7,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,7,0.010553599894046783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,15,0.012563200294971466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,15,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,15,0.010811199992895126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,15,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,15,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,15,0.010496000200510025
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,15,0.010592000186443329
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,15,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,15,0.01058880016207695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,15,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,15,0.010660800337791442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,15,0.009508799761533737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,15,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,15,0.009227199852466584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,31,0.012571200728416443
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,31,0.010782399773597717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,31,0.0126351997256279
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,31,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,31,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,31,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,31,0.01061440035700798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,31,0.01056160032749176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,31,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,31,0.010518400371074677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,31,0.010659199953079224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,31,0.010660800337791442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,31,0.010507199913263321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,31,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,63,0.012019199877977371
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,63,0.010740800201892853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,63,0.010820800065994262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,63,0.010491199791431427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,63,0.010700800269842149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,63,0.010622400045394897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,63,0.010556799918413162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,63,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,63,0.010462400317192078
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,63,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,63,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,63,0.009532800316810608
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,63,0.010630399733781815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,63,0.010662399977445603
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,127,0.01252640038728714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,127,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,127,0.01079839989542961
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,127,0.010812799632549285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,127,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,127,0.010822399705648422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,127,0.010571199655532836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,127,0.011483199894428253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,127,0.010527999699115753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,127,0.01067200005054474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,127,0.010609599947929382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,127,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,127,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,127,0.010547199845314026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,255,0.01072160005569458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,255,0.010822399705648422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,255,0.010684800148010255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,255,0.010875199735164643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,255,0.01067039966583252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,255,0.01077279970049858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,255,0.010788799822330475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,255,0.010735999792814255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,255,0.01051200032234192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,255,0.010780800133943558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,255,0.010505600273609162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,255,0.010619200021028518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,255,0.010480000078678131
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,255,0.010539200156927109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,511,0.012750400602817536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,511,0.014105600118637086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,511,0.012964800000190735
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,511,0.012744000554084778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,511,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,511,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,511,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,511,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,511,0.01071999967098236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,511,0.012606400251388549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,511,0.01080320030450821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,511,0.010903999954462052
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,511,0.01061599999666214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,511,0.01173280030488968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,1023,0.016540800034999848
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,1023,0.016771200299263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,1023,0.014912000298500061
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,1023,0.014921599626541137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,1023,0.014735999703407287
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,1023,0.014711999893188476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,1023,0.014604799449443817
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,1023,0.014627200365066529
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,1023,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,1023,0.014721600711345673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,1023,0.014643199741840363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,1023,0.012683199346065521
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,1023,0.012796799838542938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,1023,0.012736000120639801
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,2047,0.016883200407028197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,2047,0.016728000342845918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,2047,0.016729600727558136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,2047,0.014894400537014008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,2047,0.014646400511264802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,2047,0.0146479994058609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,2047,0.014718399941921234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,2047,0.01517920047044754
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,2047,0.014609600603580474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,2047,0.014572800695896148
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,2047,0.014891199767589569
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,2047,0.012796799838542938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,2047,0.012763200700283051
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,2047,0.01451839953660965
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,4095,0.02102400064468384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,4095,0.018742400407791137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,4095,0.016732800006866454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,4095,0.01669120043516159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,4095,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,4095,0.014817599952220917
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,4095,0.01475680023431778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,4095,0.018811200559139252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,4095,0.014800000190734863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,4095,0.014910399913787842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,4095,0.014851200580596923
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,4095,0.014643199741840363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,4095,0.012751999497413635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,4095,0.014667199552059173
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,8191,0.0252703994512558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,8191,0.02412319928407669
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,8191,0.01870879977941513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,8191,0.016926400363445282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,8191,0.01674560010433197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,8191,0.01674239933490753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,8191,0.016702400147914888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,8191,0.0208079993724823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,8191,0.01897920072078705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,8191,0.01685599982738495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,8191,0.016683200001716615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,8191,0.014721600711345673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,8191,0.014903999865055084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,8191,0.014779199659824372
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,16383,0.03851679861545563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,16383,0.02707839906215668
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,16383,0.02696160078048706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,16383,0.02080159932374954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,16383,0.019833600521087645
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,16383,0.018929600715637207
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,16383,0.020739200711250304
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,16383,0.022276799380779266
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,16383,0.025950399041175843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,16383,0.021091200411319733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,16383,0.018873600661754607
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,16383,0.018785600364208222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,16383,0.017518399655818938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,16383,0.018777599930763243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,32767,0.03716320097446442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,32767,0.030451199412345885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,32767,0.03307999968528748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,32767,0.028734400868415833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,32767,0.02722879946231842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,32767,0.027065598964691163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,32767,0.02712000012397766
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,32767,0.029865598678588866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,32767,0.025379198789596557
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,32767,0.024905599653720856
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,32767,0.023056000471115112
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,32767,0.02285760045051575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,32767,0.02287999987602234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,32767,0.0228752002120018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,65535,0.03737280070781708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,65535,0.03786559998989105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,65535,0.04332799911499023
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,65535,0.03463680148124695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,65535,0.03500159978866577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,65535,0.034071999788284305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,65535,0.03699199855327606
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,65535,0.0330592006444931
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,65535,0.027079999446868896
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,65535,0.03363839983940124
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,65535,0.029065600037574767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,65535,0.02712000012397766
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,65535,0.02699199914932251
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,65535,0.027193599939346315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1,1,1,131071,0.05663359761238098
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1,1,2,131071,0.058032000064849855
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1,1,4,131071,0.06831039786338806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1,1,8,131071,0.053286397457122804
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1,1,16,131071,0.05222880244255066
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1,1,32,131071,0.05339679718017578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1,1,64,131071,0.05488319993019104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1,1,1,131071,0.03928639888763428
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1,1,2,131071,0.03907999992370605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1,1,4,131071,0.039771199226379395
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1,1,8,131071,0.033139199018478394
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1,1,16,131071,0.03359360098838806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1,1,32,131071,0.03314239978790283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1,1,64,131071,0.03229759931564331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,1,0.011054400354623795
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,1,0.011432000249624253
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,1,0.010788799822330475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,1,0.010926400125026704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,1,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,1,0.010489600151777268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,1,0.010654400289058685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,1,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,1,0.010633599758148194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,1,0.010502400249242783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,1,0.01064319983124733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,1,0.010499200224876404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,1,0.010526400059461594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,1,0.010548800230026245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,3,0.012577599287033081
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,3,0.010771200060844421
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,3,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,3,0.012521600723266602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,3,0.010806400328874588
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,3,0.010652799904346467
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,3,0.010558400303125381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,3,0.010625600069761276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,3,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,3,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,3,0.0105103999376297
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,3,0.010526400059461594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,3,0.010521599650382995
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,3,0.010576000064611435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,7,0.01143840029835701
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,7,0.011523199826478958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,7,0.010883200168609618
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,7,0.010838399827480315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,7,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,7,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,7,0.010737600177526474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,7,0.010761599987745285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,7,0.010518400371074677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,7,0.010683199763298035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,7,0.010649599879980088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,7,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,7,0.010505600273609162
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,7,0.010513599961996078
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,15,0.011185599863529206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,15,0.012670400738716125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,15,0.010679999738931656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,15,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,15,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,15,0.010878399759531022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,15,0.010580799728631973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,15,0.010791999846696853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,15,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,15,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,15,0.010529600083827972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,15,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,15,0.010655999928712846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,31,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,31,0.011742399632930755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,15,0.010553599894046783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,31,0.010768000036478043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,31,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,31,0.010764800012111664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,31,0.01061279997229576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,31,0.010688000172376633
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,31,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,31,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,31,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,31,0.010655999928712846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,31,0.010777600109577179
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,31,0.010611200332641601
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,31,0.010572800040245056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,63,0.012675200402736665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,63,0.010655999928712846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,63,0.010953599959611893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,63,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,63,0.010649599879980088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,63,0.010523200035095215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,63,0.010860799998044967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,63,0.010763200372457505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,63,0.01053600013256073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,63,0.010659199953079224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,63,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,63,0.010627199709415436
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,63,0.010609599947929382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,63,0.010542400181293488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,127,0.012595200538635254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,127,0.01255359947681427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,127,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,127,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,127,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,127,0.010830400139093399
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,127,0.010583999752998351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,127,0.010737600177526474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,127,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,127,0.010966400057077408
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,127,0.010625600069761276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,127,0.010545600205659866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,127,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,127,0.010820800065994262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,255,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,255,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,255,0.010926400125026704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,255,0.011043199896812439
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,255,0.010654400289058685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,255,0.010654400289058685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,255,0.010572800040245056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,255,0.010814400017261505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,255,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,255,0.010590399801731109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,255,0.010731200128793717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,255,0.010529600083827972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,255,0.010494399815797806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,255,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,511,0.014560000598430633
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,511,0.014611199498176575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,511,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,511,0.012694400548934937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,511,0.012579199671745301
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,511,0.01281599998474121
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,511,0.012718400359153748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,511,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,511,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,511,0.012723200023174286
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,511,0.010791999846696853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,511,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,511,0.01080320030450821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,511,0.01088479980826378
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,1023,0.016785599291324615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,1023,0.014734399318695069
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,1023,0.014731200039386749
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,1023,0.014951999485492706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,1023,0.013782399892807006
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,1023,0.012849600613117218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,1023,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,1023,0.014659200608730317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,1023,0.012884800136089326
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,1023,0.012835200130939483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,1023,0.012697599828243256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,1023,0.012566399574279786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,1023,0.012619200348854064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,1023,0.012624000012874604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,2047,0.017161600291728973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,2047,0.01682559996843338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,2047,0.01480640023946762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,2047,0.014800000190734863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,2047,0.012779200077056884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,2047,0.012703999876976013
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,2047,0.014593599736690522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,2047,0.014927999675273895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,2047,0.014739200472831726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,2047,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,2047,0.012678399682044983
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,2047,0.012848000228404998
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,2047,0.012702399492263794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,2047,0.012804800271987915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,4095,0.018910400569438934
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,4095,0.0189423993229866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,4095,0.017241600155830383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,4095,0.014870400726795196
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,4095,0.014032000303268432
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,4095,0.01467359960079193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,4095,0.012827199697494508
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,4095,0.017030400037765504
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,4095,0.016654400527477263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,4095,0.016950400173664094
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,4095,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,4095,0.01276959925889969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,4095,0.012540799379348756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,4095,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,8191,0.029788801074028017
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,8191,0.021078400313854218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,8191,0.018884800374507904
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,8191,0.018742400407791137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,8191,0.01812160015106201
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,8191,0.01685439944267273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,8191,0.016841599345207216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,8191,0.022592000663280487
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,8191,0.01892479956150055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,8191,0.01686079949140549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,8191,0.016787199676036833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,8191,0.01658080071210861
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,8191,0.016748799383640288
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,16383,0.026956799626350402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,8191,0.015105600655078887
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,16383,0.03157280087471008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,16383,0.023076799511909486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,16383,0.021007999777793884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,16383,0.020972800254821778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,16383,0.020766399800777435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,16383,0.02078080028295517
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,16383,0.02592160105705261
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,16383,0.02299039959907532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,16383,0.02086080014705658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,16383,0.01896799951791763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,16383,0.018796800076961516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,16383,0.018822400271892546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,16383,0.01873600035905838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,32767,0.03212960064411163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,32767,0.035964798927307126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,32767,0.035411199927330016
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,32767,0.02730399966239929
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,32767,0.027065598964691163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,32767,0.025200000405311583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,32767,0.026979199051856993
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,32767,0.027102398872375488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,32767,0.033272001147270205
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,32767,0.02502399981021881
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,32767,0.023127999901771546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,32767,0.022937600314617158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,32767,0.02295999974012375
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,32767,0.02276960015296936
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,65535,0.050204801559448245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,65535,0.056599998474121095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,65535,0.05584319829940796
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,65535,0.051630401611328126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,65535,0.05166879892349243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,65535,0.05388000011444092
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,65535,0.053732800483703616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,65535,0.033344000577926636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,65535,0.03795199990272522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,65535,0.03344640135765076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,65535,0.030907198786735535
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,65535,0.029108801484107973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,65535,0.03067359924316406
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,65535,0.029204800724983215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,2,1,1,131071,0.07573440074920654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,2,1,2,131071,0.08280959725379944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,2,1,4,131071,0.08906400203704834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,2,1,8,131071,0.07601119875907898
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,2,1,16,131071,0.07450720071792602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,2,1,32,131071,0.07631199955940246
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,2,1,64,131071,0.07647680044174195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,2,1,1,131071,0.04955680072307587
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,2,1,2,131071,0.05414080023765564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,2,1,4,131071,0.05191199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,2,1,8,131071,0.04785760045051575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,2,1,16,131071,0.046540799736976626
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,2,1,32,131071,0.04904640018939972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,2,1,64,131071,0.04848639965057373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,1,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,1,0.012492799758911132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,1,0.012542399764060973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,1,0.01072160005569458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,1,0.010758399963378906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,1,0.01069760024547577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,1,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,1,0.012608000636100769
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,1,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,1,0.010711999982595444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,1,0.010734400153160096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,1,0.0105103999376297
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,1,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,1,0.010579200088977813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,3,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,3,0.011700800061225891
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,3,0.01282079964876175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,3,0.012718400359153748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,3,0.010907199978828431
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,3,0.010599999874830245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,3,0.012707200646400452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,3,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,3,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,3,0.01061279997229576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,3,0.010769599676132202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,3,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,3,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,3,0.010627199709415436
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,7,0.012667199969291687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,7,0.010835199803113937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,7,0.012505599856376648
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,7,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,7,0.011100800335407257
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,7,0.010603199899196624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,7,0.010558400303125381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,7,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,7,0.01080159991979599
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,7,0.010606399923563003
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,7,0.010585600137710571
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,7,0.010555200278759003
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,7,0.010673599690198899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,7,0.010608000308275222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,15,0.01268479973077774
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,15,0.010692799836397171
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,15,0.011715199798345566
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,15,0.012630400061607362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,15,0.010745599865913391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,15,0.010862399637699128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,15,0.01056319996714592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,15,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,15,0.010694400221109391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,15,0.010664000362157821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,15,0.01053759977221489
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,15,0.01053759977221489
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,15,0.010583999752998351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,15,0.010651200264692306
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,31,0.012625600397586822
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,31,0.012863999605178833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,31,0.01223680004477501
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,31,0.012326399981975555
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,31,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,31,0.010691200196743012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,31,0.010619200021028518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,31,0.012591999769210816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,31,0.010793600231409073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,31,0.01063840016722679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,31,0.010606399923563003
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,31,0.010543999820947647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,31,0.01053759977221489
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,31,0.010527999699115753
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,63,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,63,0.012583999335765839
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,63,0.011068800091743469
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,63,0.012638400495052337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,63,0.010630399733781815
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,63,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,63,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,63,0.01271039992570877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,63,0.010710400342941285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,63,0.01079839989542961
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,63,0.01051200032234192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,63,0.010620799660682679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,63,0.010497599840164185
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,63,0.01056319996714592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,127,0.012596799433231354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,127,0.012689599394798278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,127,0.010745599865913391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,127,0.010833600163459777
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,127,0.01141439974308014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,127,0.010713600367307664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,127,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,127,0.01064319983124733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,127,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,127,0.010686399787664414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,127,0.010558400303125381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,127,0.010619200021028518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,127,0.010529600083827972
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,127,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,255,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,255,0.01244639977812767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,255,0.011767999827861786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,255,0.010894399881362916
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,255,0.010740800201892853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,255,0.010903999954462052
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,255,0.010654400289058685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,255,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,255,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,255,0.010667199641466141
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,255,0.010553599894046783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,255,0.01056319996714592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,255,0.010592000186443329
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,255,0.010540799796581268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,511,0.014641599357128143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,511,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,511,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,511,0.012956799566745758
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,511,0.012889599800109864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,511,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,511,0.012593600153923034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,511,0.012830400466918945
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,511,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,511,0.012638400495052337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,511,0.012636800110340119
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,511,0.012612800300121307
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,511,0.0108815997838974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,511,0.012558400630950928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,1023,0.01698399931192398
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,1023,0.016752000153064727
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,1023,0.014817599952220917
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,1023,0.014839999377727509
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,1023,0.014711999893188476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,1023,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,1023,0.01284160017967224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,1023,0.014849600195884705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,1023,0.014603200554847717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,1023,0.013094399869441987
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,1023,0.012647999823093415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,1023,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,1023,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,1023,0.012716799974441528
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,2047,0.020815999805927278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,2047,0.01873279958963394
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,2047,0.016676799952983858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,2047,0.014841599762439728
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,2047,0.014705599844455719
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,2047,0.01465120017528534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,2047,0.012744000554084778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,2047,0.016734400391578676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,2047,0.014620800316333771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,2047,0.015046399831771851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,2047,0.012750400602817536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,2047,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,2047,0.0125231996178627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,2047,0.012868799269199371
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,4095,0.028190401196479798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,4095,0.02084160000085831
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,4095,0.018751999735832213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,4095,0.018780800700187682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,4095,0.01674399971961975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,4095,0.016774399578571318
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,4095,0.01674720048904419
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,4095,0.020819200575351714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,4095,0.016777600347995757
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,4095,0.016641600430011748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,4095,0.016651199758052827
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,4095,0.014849600195884705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,4095,0.014633600413799287
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,4095,0.014657600224018097
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,8191,0.024956800043582916
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,8191,0.029993599653244017
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,8191,0.020897600054740905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,8191,0.020470400154590607
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,8191,0.019472000002861024
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,8191,0.018771199882030486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,8191,0.018777599930763243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,8191,0.021001599729061127
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,8191,0.02285439968109131
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,8191,0.018806399405002595
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,8191,0.01685280054807663
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,8191,0.016684800386428833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,8191,0.016680000722408293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,8191,0.01668799966573715
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,16383,0.02784799933433533
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,16383,0.03120799958705902
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,16383,0.035041600465774536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,16383,0.02447360008955002
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,16383,0.02295359969139099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,16383,0.022907200455665588
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,16383,0.02197760045528412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,16383,0.025035199522972108
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,16383,0.029182401299476624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,16383,0.02340960055589676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,16383,0.02080959975719452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,16383,0.020291200280189513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,16383,0.01892800033092499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,16383,0.01886080056428909
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,32767,0.05193439722061157
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,32767,0.0531711995601654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,32767,0.0476639986038208
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,32767,0.048414400219917296
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,32767,0.04769920110702515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,32767,0.04833920001983642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,32767,0.049009600281715394
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,32767,0.03108479976654053
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,32767,0.03218240141868591
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,32767,0.031196799874305726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,32767,0.02701599895954132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,32767,0.02653760015964508
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,32767,0.025121599435806274
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,32767,0.025043201446533204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,65535,0.07115039825439454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,65535,0.0784704029560089
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,65535,0.07237759828567505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,65535,0.0712112009525299
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,65535,0.08684160113334656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,65535,0.07219359874725342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,65535,0.07254239916801453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,65535,0.04645760059356689
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,65535,0.04811840057373047
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,65535,0.04955520033836365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,65535,0.04330880045890808
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,65535,0.042059201002120974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,65535,0.04310239851474762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,65535,0.04255039989948273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,4,1,1,131071,0.12441120147705079
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,4,1,2,131071,0.13108160495758056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,4,1,4,131071,0.15384639501571656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,4,1,8,131071,0.12314879894256592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,4,1,16,131071,0.1232192039489746
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,4,1,32,131071,0.12115039825439453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,4,1,64,131071,0.12170239686965942
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,4,1,1,131071,0.0702672004699707
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,4,1,2,131071,0.07247999906539918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,4,1,4,131071,0.07889599800109863
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,4,1,8,131071,0.06509119868278504
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,4,1,16,131071,0.06492480039596557
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,4,1,32,131071,0.0652288019657135
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,4,1,64,131071,0.06453120112419128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,1,0.012625600397586822
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,1,0.012839999794960023
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,1,0.012593600153923034
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,1,0.010795199871063232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,1,0.010713600367307664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,1,0.010891199856996537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,1,0.01075040027499199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,1,0.012649600207805634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,1,0.013027200102806091
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,1,0.01085439994931221
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,1,0.010894399881362916
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,1,0.010577599704265594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,1,0.01053600013256073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,1,0.010547199845314026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,3,0.012839999794960023
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,3,0.012495999783277511
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,3,0.012127999961376191
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,3,0.011340799927711486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,3,0.011419200152158738
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,3,0.011289600282907486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,3,0.010711999982595444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,3,0.012801599502563477
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,3,0.010684800148010255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,3,0.01061440035700798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,3,0.010679999738931656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,3,0.012567999958992004
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,3,0.010582400113344192
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,3,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,7,0.012785600125789642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,7,0.012913599610328674
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,7,0.010828799754381179
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,7,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,7,0.01122559979557991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,7,0.010785599797964096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,7,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,7,0.012638400495052337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,7,0.012627199292182922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,7,0.01080320030450821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,7,0.010831999778747558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,7,0.010552000254392624
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,7,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,7,0.010639999806880952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,15,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,15,0.012639999389648438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,15,0.012491200119256973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,15,0.01263200044631958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,15,0.01080000028014183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,15,0.010681600123643876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,15,0.012838399410247803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,15,0.010814400017261505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,15,0.010870400071144103
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,15,0.010856000334024429
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,15,0.010809600353240967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,15,0.010539200156927109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,15,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,15,0.010753600299358368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,31,0.012745599448680877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,31,0.012667199969291687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,31,0.011097600311040878
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,31,0.012654399871826172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,31,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,31,0.010711999982595444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,31,0.010564800351858139
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,31,0.01276479959487915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,31,0.010828799754381179
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,31,0.010633599758148194
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,31,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,31,0.01064319983124733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,31,0.010531199723482132
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,31,0.010598400235176086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,63,0.012895999848842621
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,63,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,63,0.011297599971294403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,63,0.010704000294208527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,63,0.010804799944162368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,63,0.01143999993801117
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,63,0.010793600231409073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,63,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,63,0.011142399907112122
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,63,0.01071999967098236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,63,0.010620799660682679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,63,0.010576000064611435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,63,0.01085119992494583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,63,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,127,0.012715199589729309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,127,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,127,0.012647999823093415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,127,0.012667199969291687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,127,0.010769599676132202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,127,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,127,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,127,0.012641599774360657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,127,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,127,0.010659199953079224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,127,0.010806400328874588
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,127,0.010664000362157821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,127,0.010524799674749374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,127,0.010625600069761276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,255,0.012980799376964568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,255,0.012708799540996551
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,255,0.010675200074911118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,255,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,255,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,255,0.01074720025062561
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,255,0.010503999888896942
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,255,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,255,0.010921599715948105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,255,0.01066880002617836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,255,0.01053600013256073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,255,0.010571199655532836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,255,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,511,0.014846399426460266
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,511,0.014732800424098969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,255,0.010660800337791442
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,511,0.01467519998550415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,511,0.012835200130939483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,511,0.012761600315570831
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,511,0.012854400277137756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,511,0.012865599989891053
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,511,0.012727999687194824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,511,0.012753599882125854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,511,0.01276479959487915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,511,0.012796799838542938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,511,0.01096159964799881
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,511,0.010760000348091126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,511,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,1023,0.020897600054740905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,1023,0.01679680049419403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,1023,0.016689600050449373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,1023,0.014931200444698334
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,1023,0.014787200093269347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,1023,0.014641599357128143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,1023,0.014587199687957764
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,1023,0.01693439930677414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,1023,0.012700800597667695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,1023,0.014903999865055084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,1023,0.01467680037021637
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,1023,0.01281599998474121
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,1023,0.012750400602817536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,1023,0.012761600315570831
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,2047,0.026972800493240356
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,2047,0.021062399446964263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,2047,0.018796800076961516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,2047,0.016684800386428833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,2047,0.01698880046606064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,2047,0.01706240028142929
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,2047,0.0165583997964859
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,2047,0.020916800200939178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,2047,0.016868799924850464
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,2047,0.01488959938287735
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,2047,0.01467680037021637
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,2047,0.014630399644374847
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,2047,0.01462399959564209
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,2047,0.014660799503326416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,4095,0.02361599951982498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,4095,0.029206401109695433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,4095,0.020812800526618956
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,4095,0.019046400487422944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,4095,0.018831999599933626
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,4095,0.01687840074300766
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,4095,0.017827199399471284
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,4095,0.02128479927778244
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,4095,0.020894399285316466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,4095,0.016921600699424742
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,4095,0.016707199811935424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,4095,0.015305599570274353
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,4095,0.014769600331783294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,4095,0.014766399562358857
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,8191,0.029361599683761598
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,8191,0.030086401104927062
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,8191,0.031052801012992858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,8191,0.022278399765491487
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,8191,0.021305599808692934
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,8191,0.020715199410915375
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,8191,0.019246399402618408
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,8191,0.024889600276947022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,8191,0.025177600979804992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,8191,0.02280000001192093
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,8191,0.01879200041294098
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,8191,0.01730719953775406
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,8191,0.01793439984321594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,8191,0.016897599399089813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,16383,0.04740320146083832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,16383,0.049542400240898135
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,16383,0.05596640110015869
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,16383,0.04579519927501678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,16383,0.04552319943904877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,16383,0.04545600116252899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,16383,0.045265600085258484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,16383,0.029097598791122437
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,16383,0.02900159955024719
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,16383,0.031193599104881287
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,16383,0.024902400374412537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,16383,0.023080000281333925
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,16383,0.022974400222301482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,16383,0.02300640046596527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,32767,0.0732047975063324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,32767,0.07479519844055176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,32767,0.08851360082626343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,32767,0.07017760276794434
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,32767,0.0688048005104065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,32767,0.06945279836654664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,32767,0.0681327998638153
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,32767,0.047004801034927365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,32767,0.04603680074214935
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,32767,0.05019360184669495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,32767,0.0402895987033844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,32767,0.03974240124225616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,32767,0.03965120017528534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,32767,0.039315199851989745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,65535,0.12171360254287719
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,65535,0.12785279750823975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,65535,0.15728000402450562
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,65535,0.12140159606933594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,65535,0.120414400100708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,65535,0.12046719789505005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,65535,0.11978240013122558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,65535,0.07120000123977661
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,65535,0.06769599914550781
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,65535,0.08157280087471008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,65535,0.06378719806671143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,65535,0.06187840104103089
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,65535,0.061827200651168826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,65535,0.06235520243644714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,8,1,1,131071,0.23029119968414308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,8,1,2,131071,0.24443359375
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,8,1,4,131071,0.28956639766693115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,8,1,8,131071,0.2357680082321167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,8,1,16,131071,0.23061439990997315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,8,1,32,131071,0.23487679958343505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,8,1,64,131071,0.23685760498046876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,8,1,1,131071,0.1249951958656311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,8,1,2,131071,0.11457439661026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,8,1,4,131071,0.1420207977294922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,8,1,8,131071,0.10479680299758912
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,8,1,16,131071,0.10413600206375122
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,8,1,32,131071,0.10415199995040894
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,8,1,64,131071,0.10382879972457885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,1,0.014737600088119506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,1,0.012811200320720672
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,1,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,1,0.012647999823093415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,1,0.010728000104427338
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,1,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,1,0.01263359934091568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,1,0.012873600423336028
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,1,0.012729600071907043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,1,0.012611199915409089
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,1,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,1,0.010628800094127654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,1,0.010545600205659866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,1,0.010702399909496308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,3,0.014795200526714325
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,3,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,3,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,3,0.01268800050020218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,3,0.012548799812793731
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,3,0.010468800365924836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,3,0.01074879989027977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,3,0.012907199561595917
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,3,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,3,0.01257600039243698
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,3,0.010657600313425063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,3,0.01080000028014183
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,3,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,3,0.01058719977736473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,7,0.014692799746990204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,7,0.012777599692344665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,7,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,7,0.010817600041627884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,7,0.011030399799346923
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,7,0.01064160019159317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,7,0.011268799751996994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,7,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,7,0.012689599394798278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,7,0.012627199292182922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,7,0.010683199763298035
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,7,0.010596799850463866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,7,0.010583999752998351
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,7,0.01063840016722679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,15,0.01467519998550415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,15,0.012716799974441528
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,15,0.012838399410247803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,15,0.01075040027499199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,15,0.012628799676895142
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,15,0.010796800255775452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,15,0.011761599779129028
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,15,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,15,0.012572799623012543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,15,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,15,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,15,0.010486400127410889
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,15,0.01080320030450821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,15,0.0106175996363163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,31,0.014723199605941772
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,31,0.012734399735927581
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,31,0.012761600315570831
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,31,0.011764799803495407
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,31,0.012249600142240524
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,31,0.010657600313425063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,31,0.011153600364923476
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,31,0.012763200700283051
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,31,0.012676799297332763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,31,0.012548799812793731
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,31,0.010814400017261505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,31,0.010567999631166457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,31,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,31,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,63,0.014839999377727509
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,63,0.012803199887275695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,63,0.012611199915409089
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,63,0.012392000108957291
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,63,0.012484800070524216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,63,0.010790400207042694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,63,0.012516799569129943
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,63,0.012699200212955475
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,63,0.012574400007724761
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,63,0.011371199786663056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,63,0.010593599826097488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,63,0.01053439974784851
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,63,0.01072160005569458
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,63,0.010662399977445603
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,127,0.014776000380516052
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,127,0.012734399735927581
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,127,0.012779200077056884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,127,0.010707200318574906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,127,0.012532800436019897
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,127,0.010681600123643876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,127,0.011633600294589996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,127,0.012656000256538392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,127,0.012558400630950928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,127,0.010704000294208527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,127,0.010574399679899215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,127,0.012100800126791
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,127,0.010676799714565277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,127,0.010606399923563003
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,255,0.0146479994058609
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,255,0.0133200004696846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,255,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,255,0.012606400251388549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,255,0.010739199817180634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,255,0.010742399841547012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,255,0.010791999846696853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,255,0.012681600451469422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,255,0.012622399628162384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,255,0.012563200294971466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,255,0.010817600041627884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,255,0.010790400207042694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,255,0.010652799904346467
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,255,0.010654400289058685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,511,0.01882079988718033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,511,0.016732800006866454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,511,0.01481119990348816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,511,0.01467359960079193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,511,0.014721600711345673
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,511,0.014705599844455719
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,511,0.012702399492263794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,511,0.014750400185585022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,511,0.014654399454593658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,511,0.012849600613117218
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,511,0.01255200058221817
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,511,0.011558400094509124
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,511,0.012579199671745301
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,511,0.012639999389648438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,1023,0.024879999458789825
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,1023,0.020822399854660036
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,1023,0.017603200674057008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,1023,0.016641600430011748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,1023,0.016795200109481812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,1023,0.016675199568271636
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,1023,0.018889600038528444
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,1023,0.01682399958372116
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,1023,0.015364800393581391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,1023,0.014847999811172486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,1023,0.01478080004453659
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,1023,0.012758399546146392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,1023,0.012836800515651703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,1023,0.012768000364303589
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,2047,0.022414399683475493
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,2047,0.028913599252700806
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,2047,0.02094399929046631
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,2047,0.018873600661754607
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,2047,0.016828800737857818
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,2047,0.016908800601959227
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,2047,0.01879359930753708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,2047,0.02083680033683777
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,2047,0.020972800254821778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,2047,0.01679359972476959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,2047,0.01485760062932968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,2047,0.014737600088119506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,2047,0.014740799367427827
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,2047,0.014731200039386749
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,4095,0.029099199175834655
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,4095,0.029124799370765685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,4095,0.029065600037574767
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,4095,0.020871999859809875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,4095,0.020795199275016784
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,4095,0.020870399475097657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,4095,0.018935999274253844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,4095,0.023204800486564637
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,4095,0.023035199940204622
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,4095,0.020838400721549986
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,4095,0.01687040030956268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,4095,0.016711999475955964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,4095,0.01684959977865219
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,4095,0.01672160029411316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,8191,0.04758560061454773
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,8191,0.05009920001029968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,8191,0.053457599878311154
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,8191,0.045351999998092654
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,8191,0.04492000043392182
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,8191,0.04381119906902313
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,8191,0.0438832014799118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,8191,0.02987839877605438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,8191,0.029068800806999206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,8191,0.029084798693656922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,8191,0.022840000689029694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,8191,0.022360000014305114
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,8191,0.02102559953927994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,8191,0.0210207998752594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,16383,0.07523519992828369
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,16383,0.07512959837913513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,16383,0.0900816023349762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,16383,0.07246879935264587
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,16383,0.07303040027618408
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,16383,0.07194399833679199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,16383,0.07085279822349548
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,16383,0.047486400604248045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,16383,0.045747199654579164
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,16383,0.04968799948692322
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,16383,0.039585599303245546
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,16383,0.039284801483154295
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,16383,0.039396798610687254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,16383,0.03948160111904144
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,32767,0.12719199657440186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,32767,0.12507359981536864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,32767,0.16274399757385255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,32767,0.12126239538192748
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,32767,0.12153279781341553
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,32767,0.12009600400924683
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,32767,0.12151520252227783
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,32767,0.07424479722976685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,32767,0.06832159757614135
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,32767,0.08379359841346741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,32767,0.06271839737892151
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,32767,0.062097597122192386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,32767,0.06139519810676575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,32767,0.06157119870185852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,65535,0.23579840660095214
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,65535,0.22953441143035888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,65535,0.3087935924530029
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,65535,0.21916639804840088
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,65535,0.21716639995574952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,65535,0.21858720779418944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,65535,0.21704320907592772
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,65535,0.13527040481567382
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,65535,0.11385760307312012
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,65535,0.15106240510940552
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,65535,0.10537120103836059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,65535,0.10487680435180664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,65535,0.10447039604187011
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,65535,0.10494719743728638
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,16,1,1,131071,0.44582719802856446
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,16,1,2,131071,0.4347599983215332
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,16,1,4,131071,0.6017024040222168
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,16,1,8,131071,0.4213696002960205
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,16,1,16,131071,0.419920015335083
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,16,1,32,131071,0.4198592185974121
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,16,1,64,131071,0.4207183837890625
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,16,1,1,131071,0.24053759574890138
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,16,1,2,131071,0.20862240791320802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,16,1,4,131071,0.28672959804534914
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,16,1,8,131071,0.19046720266342163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,16,1,16,131071,0.19098559617996216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,16,1,32,131071,0.19124319553375244
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,16,1,64,131071,0.19025280475616455
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,1,0.014692799746990204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,1,0.01462240070104599
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,1,0.012824000418186187
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,1,0.012812800705432892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,1,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,1,0.012542399764060973
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,1,0.012644800543785095
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,1,0.012788799405097962
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,1,0.014668799936771393
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,1,0.012651200592517852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,1,0.012683199346065521
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,1,0.011020799726247787
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,1,0.012580800056457519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,1,0.01071999967098236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,3,0.014703999459743499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,3,0.014748799800872802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,3,0.01279039978981018
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,3,0.012544000148773193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,3,0.012636800110340119
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,3,0.012643200159072877
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,3,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,3,0.012630400061607362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,3,0.013020800054073333
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,3,0.012806400656700134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,3,0.011156799644231797
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,3,0.011796800047159195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,3,0.010843200236558914
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,3,0.011084800213575363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,7,0.014739200472831726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,7,0.014619199931621552
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,7,0.012828800082206725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,7,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,7,0.012556800246238708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,7,0.012556800246238708
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,7,0.012664000689983367
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,7,0.014707200229167938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,7,0.012836800515651703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,7,0.012624000012874604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,7,0.012705600261688233
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,7,0.01079839989542961
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,7,0.010945600271224976
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,7,0.010604800283908844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,15,0.01462559998035431
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,15,0.014696000516414643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,15,0.01276479959487915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,15,0.012614400684833526
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,15,0.012796799838542938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,15,0.012673600018024445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,15,0.012680000066757202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,15,0.012723200023174286
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,15,0.012830400466918945
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,15,0.012572799623012543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,15,0.012612800300121307
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,15,0.010713600367307664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,15,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,15,0.010684800148010255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,31,0.014814400672912597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,31,0.014636799693107605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,31,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,31,0.012774400413036346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,31,0.012595200538635254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,31,0.012700800597667695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,31,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,31,0.014710399508476257
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,31,0.01273760050535202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,31,0.012771199643611907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,31,0.012624000012874604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,31,0.010790400207042694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,31,0.010740800201892853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,31,0.010648000240325927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,63,0.01462240070104599
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,63,0.01478399932384491
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,63,0.012902399897575379
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,63,0.012748800218105316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,63,0.012651200592517852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,63,0.0127920001745224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,63,0.012574400007724761
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,63,0.012742400169372559
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,63,0.01276479959487915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,63,0.012678399682044983
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,63,0.010806400328874588
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,63,0.010820800065994262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,63,0.010708799958229065
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,63,0.010849600285291671
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,127,0.013044799864292144
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,127,0.014574399590492249
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,127,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,127,0.012875199317932129
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,127,0.012619200348854064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,127,0.012649600207805634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,127,0.012697599828243256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,127,0.012860800325870513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,127,0.012856000661849975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,127,0.012627199292182922
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,127,0.011868800222873687
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,127,0.01072480008006096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,127,0.010608000308275222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,127,0.010755199939012527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,255,0.014870400726795196
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,255,0.014774399995803832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,255,0.013820800185203552
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,255,0.012807999551296235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,255,0.01268479973077774
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,255,0.012617599964141846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,255,0.012777599692344665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,255,0.014772799611091614
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,255,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,255,0.012678399682044983
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,255,0.011095999926328658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,255,0.010718400031328202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,255,0.010601600259542465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,255,0.010896000266075134
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,511,0.02095839977264404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,511,0.018743999302387238
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,511,0.016888000071048737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,511,0.014787200093269347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,511,0.014833599328994751
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,511,0.014744000136852264
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,511,0.014750400185585022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,511,0.020852799713611602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,511,0.014720000326633453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,511,0.013014400005340576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,511,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,511,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,511,0.012671999633312225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,511,0.012567999958992004
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,1023,0.024372799694538115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,1023,0.024982400238513947
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,1023,0.02088959962129593
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,1023,0.016974399983882903
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,1023,0.01703519970178604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,1023,0.01684480011463165
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,1023,0.01684959977865219
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,1023,0.022841599583625794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,1023,0.0189983993768692
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,1023,0.016812799870967864
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,1023,0.014774399995803832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,1023,0.014575999975204468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,1023,0.014769600331783294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,1023,0.014660799503326416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,2047,0.031038400530815125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,2047,0.026979199051856993
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,2047,0.031113600730895995
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,2047,0.0208624005317688
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,2047,0.020742399990558623
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,2047,0.02051520049571991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,2047,0.020916800200939178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,2047,0.02507520020008087
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,2047,0.021036800742149354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,2047,0.016840000450611115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,2047,0.02086080014705658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,2047,0.014932799339294433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,2047,0.014870400726795196
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,2047,0.014843200147151948
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,4095,0.05101760029792786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,4095,0.0497408002614975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,4095,0.05419840216636658
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,4095,0.04323999881744385
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,4095,0.04262399971485138
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,4095,0.04174399971961975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,4095,0.04139040112495422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,4095,0.03313280045986176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,4095,0.02736000120639801
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,4095,0.02794879972934723
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,4095,0.02093919962644577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,4095,0.02044160068035126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,4095,0.020796799659729005
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,4095,0.01889919936656952
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,8191,0.07672799825668335
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,8191,0.07726240158081055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,8191,0.09143360257148743
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,8191,0.0701919972896576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,8191,0.06988160014152527
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,8191,0.07085760235786438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,8191,0.06892960071563721
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,8191,0.050539201498031615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,8191,0.04551520049571991
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,8191,0.0513264000415802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,8191,0.03933599889278412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,8191,0.03700959980487824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,8191,0.03722079992294312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,8191,0.03720960021018982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,16383,0.13300319910049438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,16383,0.1283792018890381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,16383,0.16420960426330566
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,16383,0.12162239551544189
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,16383,0.11941440105438232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,16383,0.11929600238800049
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,16383,0.11955679655075073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,16383,0.07992640137672424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,16383,0.06993280053138733
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,16383,0.08419839739799499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,16383,0.06049119830131531
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,16383,0.06004959940910339
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,16383,0.05933279991149902
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,16383,0.05889279842376709
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,32767,0.23720319271087648
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,32767,0.23408000469207763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,32767,0.31120479106903076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,32767,0.22769598960876464
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,32767,0.22387518882751464
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,32767,0.22504959106445313
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,32767,0.2239232063293457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,32767,0.13495839834213258
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,32767,0.11910560131072997
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,32767,0.1524656057357788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,32767,0.10475679636001586
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,32767,0.10593600273132324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,32767,0.1046720027923584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,32767,0.10531040430068969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,32,1,2,65535,0.4397088050842285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,32,1,1,65535,0.4567584037780762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,32,1,4,65535,0.6034976005554199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,32,1,8,65535,0.4317215919494629
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,32,1,32,65535,0.4301328182220459
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,32,1,16,65535,0.4288496017456055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,32,1,64,65535,0.4345407962799072
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,32,1,1,65535,0.2546128034591675
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,32,1,2,65535,0.2142335891723633
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,32,1,4,65535,0.2857183933258057
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,32,1,8,65535,0.19531840085983276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,32,1,16,65535,0.1942064046859741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,32,1,32,65535,0.19311840534210206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,1,0.01621599942445755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,1,0.014761599898338317
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,32,1,64,65535,0.19275840520858764
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,1,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,1,0.012748800218105316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,1,0.01268640011548996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,1,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,1,0.012768000364303589
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,1,0.015889599919319153
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,1,0.012814399600028992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,1,0.014689600467681885
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,1,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,1,0.012646399438381195
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,1,0.012615999579429627
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,1,0.012628799676895142
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,3,0.01613440066576004
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,3,0.014640000462532044
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,3,0.014696000516414643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,3,0.012751999497413635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,3,0.012479999661445617
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,3,0.012859199941158295
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,3,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,3,0.014886400103569031
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,3,0.012521600723266602
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,3,0.012731200456619263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,3,0.01072319969534874
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,3,0.012588800489902496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,3,0.010716799646615982
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,3,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,7,0.014844800531864166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,7,0.014726400375366211
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,7,0.014678399264812469
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,7,0.012740799784660339
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,7,0.013212800025939941
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,7,0.01279360055923462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,7,0.012624000012874604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,7,0.014854399859905243
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,7,0.012763200700283051
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,7,0.012591999769210816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,7,0.01282079964876175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,7,0.011510399729013443
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,7,0.010676799714565277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,7,0.010766399651765823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,15,0.016761599481105803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,15,0.014750400185585022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,15,0.01467519998550415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,15,0.013219200074672699
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,15,0.01290079951286316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,15,0.012814399600028992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,15,0.012665599584579468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,15,0.014884799718856812
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,15,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,15,0.012774400413036346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,15,0.012587200105190276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,15,0.010524799674749374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,15,0.010871999710798264
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,15,0.011500799655914306
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,31,0.014814400672912597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,31,0.014753599464893342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,31,0.014708800613880158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,31,0.014716799557209014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,31,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,31,0.01273760050535202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,31,0.012759999930858612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,31,0.016663999855518342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,31,0.01282079964876175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,31,0.012881599366664886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,31,0.012572799623012543
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,31,0.010744000226259232
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,31,0.010751999914646149
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,31,0.010841599851846694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,63,0.01496960073709488
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,63,0.013950400054454803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,63,0.014763200283050537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,63,0.013372799754142762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,63,0.01281760036945343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,63,0.012604799866676331
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,63,0.012857599556446076
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,63,0.014975999295711518
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,63,0.012980799376964568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,63,0.0127920001745224
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,63,0.01266240030527115
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,63,0.012571200728416443
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,63,0.012540799379348756
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,63,0.010635200142860412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,127,0.015297600626945495
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,127,0.01472959965467453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,127,0.014763200283050537
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,127,0.012783999741077422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,127,0.012811200320720672
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,127,0.012835200130939483
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,127,0.012732799351215362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,127,0.01499200016260147
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,127,0.012729600071907043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,127,0.01287200003862381
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,127,0.012692800164222718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,127,0.010636799782514573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,127,0.010726399719715118
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,127,0.010822399705648422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,255,0.01695519983768463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,255,0.014772799611091614
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,255,0.014766399562358857
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,255,0.012966400384902954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,255,0.012803199887275695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,255,0.012580800056457519
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,255,0.012779200077056884
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,255,0.016774399578571318
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,255,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,255,0.01271360069513321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,255,0.012654399871826172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,255,0.011035200208425522
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,255,0.010862399637699128
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,255,0.010702399909496308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,511,0.020803199708461763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,511,0.02309280037879944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,511,0.01873279958963394
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,511,0.016711999475955964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,511,0.01669600009918213
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,511,0.01501920074224472
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,511,0.016966399550437928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,511,0.018787199258804323
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,511,0.014894400537014008
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,511,0.02094080001115799
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,511,0.014703999459743499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,511,0.012811200320720672
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,511,0.012804800271987915
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,511,0.012753599882125854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,1023,0.026966398954391478
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,1023,0.0311024010181427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,1023,0.029396799206733704
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,1023,0.02106720060110092
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,1023,0.018881599605083465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,1023,0.018753600120544434
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,1023,0.01886720061302185
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,1023,0.02099999934434891
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,1023,0.02340960055589676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,1023,0.01879040002822876
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,1023,0.016923199594020843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,1023,0.01478240042924881
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,1023,0.014983999729156493
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,1023,0.01454080045223236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,2047,0.046132799983024594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,2047,0.05201759934425354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,2047,0.05177119970321655
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,2047,0.04374400079250336
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,2047,0.043463999032974245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,2047,0.04179359972476959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,2047,0.04272960126399994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,2047,0.028380799293518066
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,2047,0.029104000329971312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,2047,0.02728799879550934
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,2047,0.021140800416469575
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,2047,0.0188960000872612
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,2047,0.018799999356269838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,2047,0.018855999410152435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,4095,0.07332320213317871
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,4095,0.07947679758071899
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,4095,0.08845919966697693
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,4095,0.07233279943466187
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,4095,0.07033439874649047
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,4095,0.0680895984172821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,4095,0.06996319890022278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,4095,0.046140798926353456
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,4095,0.04943360090255737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,4095,0.04857119917869568
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,4095,0.03940640091896057
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,4095,0.03750720024108887
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,4095,0.03590719997882843
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,4095,0.03722560107707977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,8191,0.1262943983078003
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,8191,0.13143680095672608
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,8191,0.1619328022003174
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,8191,0.12242560386657715
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,8191,0.12117120027542114
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,8191,0.12114559412002564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,8191,0.11989279985427856
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,8191,0.07320320010185241
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,8191,0.07319039702415467
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,8191,0.08244640231132508
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,8191,0.06324959993362426
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,8191,0.060335999727249144
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,8191,0.05986080169677734
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,8191,0.05994240045547485
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,16383,0.23766400814056396
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,16383,0.23608479499816895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,16383,0.30541601181030276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,16383,0.22891199588775635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,16383,0.22645280361175538
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,16383,0.22528319358825682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,16383,0.22379519939422607
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,16383,0.12752799987792968
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,16383,0.12305599451065063
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,16383,0.15068960189819336
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,16383,0.10733439922332763
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,16383,0.10478719472885131
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,16383,0.10535839796066285
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,16383,0.10544159412384033
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,64,1,1,32767,0.44930081367492675
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,64,1,2,32767,0.442414379119873
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,64,1,4,32767,0.5938047885894775
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,64,1,8,32767,0.43248639106750486
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,64,1,16,32767,0.4361231803894043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,64,1,32,32767,0.4296288013458252
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,64,1,64,32767,0.4302031993865967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,64,1,1,32767,0.24028959274291992
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,64,1,2,32767,0.21744959354400634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,64,1,4,32767,0.2835279941558838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,64,1,8,32767,0.19750880002975463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,64,1,16,32767,0.19512640237808226
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,64,1,32,32767,0.1939679980278015
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,64,1,64,32767,0.19353920221328735
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,1,0.021060800552368163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,1,0.01685439944267273
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,1,0.01990240067243576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,1,0.014740799367427827
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,1,0.01467359960079193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,1,0.013459199666976928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,1,0.014748799800872802
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,1,0.02099999934434891
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,1,0.016735999286174773
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,1,0.012812800705432892
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,1,0.017392000555992125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,1,0.012624000012874604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,1,0.012671999633312225
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,1,0.012603199481964112
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,3,0.02109760046005249
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,3,0.01669120043516159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,3,0.020880000293254854
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,3,0.014735999703407287
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,3,0.014550399780273438
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,3,0.014627200365066529
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,3,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,3,0.021003200113773345
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,3,0.016761599481105803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,3,0.01706400066614151
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,3,0.012777599692344665
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,3,0.012583999335765839
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,3,0.012204799801111221
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,7,0.02093919962644577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,3,0.01271200031042099
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,7,0.01679999977350235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,7,0.019300800561904908
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,7,0.014839999377727509
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,7,0.014734399318695069
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,7,0.014641599357128143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,7,0.014679999649524688
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,7,0.02287680059671402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,7,0.016803200542926788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,7,0.0186256006360054
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,7,0.012780800461769104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,7,0.012785600125789642
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,7,0.012647999823093415
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,7,0.012680000066757202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,15,0.020870399475097657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,15,0.016927999258041383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,15,0.020791999995708466
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,15,0.01481119990348816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,15,0.014691199362277984
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,15,0.014641599357128143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,15,0.014734399318695069
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,15,0.021668800711631776
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,15,0.014900800585746766
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,15,0.017580799758434296
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,15,0.012755200266838074
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,15,0.012590399384498597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,15,0.012630400061607362
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,15,0.012700800597667695
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,31,0.021070399880409242
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,31,0.016740800440311433
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,31,0.02080959975719452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,31,0.014796799421310425
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,31,0.014688000082969666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,31,0.014697599411010741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,31,0.014936000108718872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,31,0.021211199462413788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,31,0.016731199622154237
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,31,0.017636799812316896
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,31,0.012772800028324127
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,31,0.01255359947681427
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,31,0.012668800354003907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,31,0.012564800679683685
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,63,0.021006399393081666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,63,0.01679359972476959
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,63,0.020795199275016784
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,63,0.01472959965467453
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,63,0.014636799693107605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,63,0.014672000706195832
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,63,0.01334560066461563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,63,0.020921599864959717
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,63,0.016803200542926788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,63,0.017854399979114532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,63,0.012723200023174286
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,63,0.012691199779510498
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,63,0.012651200592517852
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,63,0.012729600071907043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,127,0.020678399503231047
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,127,0.016846400499343873
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,127,0.02080159932374954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,127,0.014769600331783294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,127,0.0147024005651474
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,127,0.014604799449443817
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,127,0.014918400347232819
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,127,0.021003200113773345
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,127,0.016654400527477263
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,127,0.01767680048942566
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,127,0.012908799946308136
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,127,0.01252640038728714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,127,0.012622399628162384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,127,0.01266079992055893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,255,0.0269679993391037
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,255,0.018875199556350707
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,255,0.020897600054740905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,255,0.01486240029335022
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,255,0.014804799854755402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,255,0.014575999975204468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,255,0.014105600118637086
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,255,0.022974400222301482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,255,0.016926400363445282
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,255,0.01807039976119995
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,255,0.01281919926404953
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,255,0.012598399817943574
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,255,0.01265760064125061
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,255,0.013062399625778199
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,511,0.03600319921970367
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,511,0.025920000672340394
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,511,0.03135200142860413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,511,0.018905599415302277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,511,0.0182096004486084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,511,0.017748799920082093
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,511,0.01879200041294098
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,511,0.02727360129356384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,511,0.0187376007437706
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,511,0.020926399528980254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,511,0.014953599870204925
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,511,0.014727999269962311
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,511,0.014638400077819825
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,511,0.014732800424098969
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,1023,0.05171520113945007
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,1023,0.04944159984588623
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,1023,0.053814399242401126
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,1023,0.04237920045852661
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,1023,0.04125120043754578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,1023,0.04105600118637085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,1023,0.03962239921092987
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,1023,0.037278398871421814
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,1023,0.025095999240875244
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,1023,0.03096640110015869
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,1023,0.018769599497318268
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,1023,0.016952000558376312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,1023,0.01881760060787201
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,1023,0.016897599399089813
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,2047,0.08031520247459412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,2047,0.07495679855346679
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,2047,0.09348000288009643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,2047,0.06835839748382569
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,2047,0.06733760237693787
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,2047,0.06644639968872071
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,2047,0.06615039706230164
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,2047,0.053144001960754396
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,2047,0.04556480050086975
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,2047,0.05139039754867554
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,2047,0.037457600235939026
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,2047,0.036348798871040346
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,2047,0.03578400015830994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,2047,0.035329601168632506
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,4095,0.13621280193328858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,4095,0.12701599597930907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,4095,0.1673743963241577
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,4095,0.12099519968032837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,4095,0.11894559860229492
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,4095,0.11775519847869872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,4095,0.11863360404968262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,4095,0.0807263970375061
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,4095,0.07051360011100768
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,4095,0.08703839778900146
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,4095,0.06055999994277954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,4095,0.05861759781837463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,4095,0.05961120128631592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,4095,0.05890880227088928
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,8191,0.24268639087677002
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,8191,0.23214399814605713
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,8191,0.22301759719848632
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,8191,0.31391520500183107
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,8191,0.2233407974243164
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,8191,0.22188000679016112
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,8191,0.22036960124969482
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,8191,0.14320000410079955
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,8191,0.11896480321884155
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,8191,0.15438719987869262
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,8191,0.10643999576568604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,8191,0.10521279573440552
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,8191,0.10394560098648072
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,8191,0.10440959930419921
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,128,1,1,16383,0.45435681343078616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,128,1,2,16383,0.4390751838684082
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,128,1,4,16383,0.603984022140503
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,128,1,8,16383,0.431660795211792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,128,1,16,16383,0.4263152122497559
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,128,1,32,16383,0.42516322135925294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,128,1,64,16383,0.4228015899658203
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,128,1,1,16383,0.25155999660491946
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,128,1,2,16383,0.21485600471496583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,128,1,4,16383,0.28876960277557373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,128,1,8,16383,0.19509439468383788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,128,1,16,16383,0.1933840036392212
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,128,1,32,16383,0.19334080219268798
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,1,0.03611679971218109
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,128,1,64,16383,0.1934191942214966
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,1,0.024886399507522583
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,1,0.020712000131607056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,1,0.01887039989233017
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,1,0.031119999289512635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,1,0.01875839978456497
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,1,0.018935999274253844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,1,0.0352975994348526
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,1,0.022998400032520294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,1,0.02545759975910187
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,1,0.018212799727916718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,1,0.016652800142765045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,1,0.016187199950218202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,1,0.015137599408626556
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,3,0.03583039939403534
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,3,0.024846400320529937
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,3,0.031091201305389404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,3,0.020772799849510193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,3,0.01887679994106293
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,3,0.019032000005245207
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,3,0.019257600605487823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,3,0.03533760011196137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,3,0.023024000227451324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,3,0.02704159915447235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,3,0.01838400065898895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,3,0.01688639968633652
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,3,0.016784000396728515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,3,0.016728000342845918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,7,0.03731519877910614
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,7,0.024903999269008638
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,7,0.031091201305389404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,7,0.020883199572563172
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,7,0.019246399402618408
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,7,0.018929600715637207
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,7,0.019441600143909454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,7,0.0353520005941391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,7,0.022937600314617158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,7,0.025003200769424437
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,7,0.018702399730682374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,7,0.016803200542926788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,7,0.016417600214481354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,7,0.016468800604343414
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,15,0.03718239963054657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,15,0.02492479979991913
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,15,0.03115679919719696
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,15,0.020828799903392793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,15,0.019750399887561797
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,15,0.019355200231075287
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,15,0.01902559995651245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,15,0.035497599840164186
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,15,0.022976000607013703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,15,0.02682879865169525
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,15,0.017740799486637114
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,15,0.0169295996427536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,15,0.016755199432373045
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,15,0.016785599291324615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,31,0.0372655987739563
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,31,0.0249551996588707
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,31,0.031176000833511353
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,31,0.02088479995727539
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,31,0.02011999934911728
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,31,0.020601600408554077
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,31,0.01892320066690445
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,31,0.035364800691604616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,31,0.026188799738883974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,31,0.0186271995306015
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,31,0.022912000119686127
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,31,0.016683200001716615
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,31,0.016828800737857818
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,31,0.016784000396728515
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,63,0.03755680024623871
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,63,0.02487040013074875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,63,0.031244799494743347
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,63,0.020900799334049223
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,63,0.019233599305152893
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,63,0.01926079988479614
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,63,0.019679999351501463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,63,0.037169599533081056
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,63,0.023051199316978455
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,63,0.025231999158859254
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,63,0.018721599876880646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,63,0.016859200596809388
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,63,0.016820800304412842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,63,0.016129599511623384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,127,0.0378928005695343
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,127,0.02576960027217865
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,127,0.031164801120758055
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,127,0.020815999805927278
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,127,0.019726400077342988
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,127,0.019566400349140166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,127,0.019806399941444397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,127,0.03561759889125824
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,127,0.022753599286079406
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,127,0.026977598667144775
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,127,0.016814400255680085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,127,0.01674560010433197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,127,0.018796800076961516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,127,0.01674560010433197
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,255,0.04714080095291138
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,255,0.03731679916381836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,255,0.035787200927734374
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,255,0.022510400414466857
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,255,0.0195360004901886
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,255,0.019704000651836397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,255,0.019409599900245666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,255,0.04137920141220093
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,255,0.027107200026512145
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,255,0.025155198574066163
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,255,0.018799999356269838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,255,0.017688000202178956
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,255,0.016254399716854096
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,255,0.016752000153064727
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,511,0.06175839900970459
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,511,0.054028797149658206
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,511,0.058064001798629764
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,511,0.04374879896640778
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,511,0.04214560091495514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,511,0.04166879951953888
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,511,0.041489601135253906
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,511,0.04975839853286743
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,511,0.0379584014415741
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,511,0.036620798707008365
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,511,0.022566400468349457
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,511,0.020870399475097657
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,511,0.01913439929485321
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,511,0.0198512002825737
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,1023,0.09082400202751159
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,1023,0.08226879835128784
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,1023,0.09616159796714782
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,1023,0.07093759775161743
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,1023,0.0683456003665924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,1023,0.06812480092048645
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,1023,0.06744639873504639
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,1023,0.06471359729766846
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,1023,0.052007997035980226
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,1023,0.05538880228996277
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,1023,0.03978239893913269
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,1023,0.03834879994392395
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,1023,0.037334400415420535
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,1023,0.0373663991689682
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,2047,0.1463503956794739
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,2047,0.13867839574813842
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,2047,0.17123359441757202
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,2047,0.12481600046157837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,2047,0.12187360525131226
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,2047,0.12168480157852173
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,2047,0.12141920328140259
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,2047,0.09304800033569335
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,2047,0.07825120091438294
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,2047,0.09054719805717468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,2047,0.06356160044670105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,2047,0.06104320287704468
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,2047,0.06032480001449585
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,2047,0.060382401943206786
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,4095,0.2562495946884155
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,4095,0.244979190826416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,4095,0.31733920574188235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,4095,0.22918241024017333
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,4095,0.2271023988723755
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,4095,0.2261359930038452
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,4095,0.2233680009841919
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,4095,0.152292799949646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,4095,0.12744799852371216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,4095,0.15939199924468994
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,4095,0.1109760046005249
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,4095,0.10626239776611328
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,4095,0.10627199411392212
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,4095,0.10537120103836059
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,256,1,1,8191,0.48694558143615724
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,256,1,2,8191,0.45194401741027834
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,256,1,4,8191,0.6057407855987549
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,256,1,8,8191,0.433673620223999
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,256,1,16,8191,0.43296160697937014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,256,1,32,8191,0.43087201118469237
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,256,1,64,8191,0.4307263851165771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,256,1,1,8191,0.27176799774169924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,256,1,2,8191,0.22637760639190674
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,256,1,4,8191,0.29562718868255616
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,256,1,8,8191,0.19935200214385987
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,256,1,16,8191,0.1972383975982666
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,256,1,32,8191,0.1968400001525879
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,256,1,64,8191,0.19566719532012938
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,1,0.06352800130844116
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,1,0.04232319891452789
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,1,0.02908160090446472
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,1,0.05054240226745606
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,1,0.03109920024871826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,1,0.02913280129432678
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,1,0.029193601012229918
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,1,0.037376001477241516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,1,0.06381120085716248
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,1,0.04209440052509308
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,1,0.027024000883102417
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,1,0.022995199263095855
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,1,0.023039999604225158
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,1,0.023027199506759643
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,3,0.06358720064163208
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,3,0.04337120056152344
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,3,0.051420801877975465
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,3,0.03134399950504303
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,3,0.029126399755477907
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,3,0.029104000329971312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,3,0.029198399186134337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,3,0.06399999856948853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,3,0.03739840090274811
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,3,0.04316000044345856
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,3,0.026947200298309326
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,3,0.023035199940204622
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,3,0.022932800650596618
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,3,0.023124800622463228
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,7,0.043350398540496826
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,7,0.06397280097007751
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,7,0.04983839988708496
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,7,0.03296639919281006
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,7,0.029204800724983215
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,7,0.028972798585891725
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,7,0.029139199852943422
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,7,0.06376799941062927
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,7,0.03789759874343872
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,7,0.02714880108833313
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,7,0.04148159921169281
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,7,0.023228800296783446
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,7,0.023148800432682037
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,7,0.023078399896621703
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,15,0.06402080059051514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,15,0.043356800079345705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,15,0.05063520073890686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,15,0.031196799874305726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,15,0.028939199447631837
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,15,0.02922559976577759
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,15,0.029216000437736513
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,15,0.06386719942092896
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,15,0.03920480012893677
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,15,0.0423440009355545
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,15,0.027153599262237548
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,15,0.023134399950504304
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,15,0.02290080040693283
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,15,0.023183999955654143
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,31,0.06417120099067689
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,31,0.04364959895610809
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,31,0.04973439872264862
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,31,0.03148640096187592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,31,0.02919999957084656
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,31,0.029131200909614564
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,31,0.029049599170684816
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,31,0.06389759778976441
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,31,0.03935680091381073
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,31,0.04180479943752289
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,31,0.026974400877952574
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,31,0.023175999522209167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,31,0.023105600476264955
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,31,0.022840000689029694
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,63,0.0642192006111145
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,63,0.045335999131202696
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,63,0.051806402206420896
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,63,0.03172479867935181
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,63,0.029054400324821473
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,63,0.02924000024795532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,63,0.029187199473381043
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,63,0.0639136016368866
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,63,0.039719998836517334
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,63,0.042735999822616576
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,63,0.026976001262664796
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,63,0.023000000417232512
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,63,0.023048000037670137
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,63,0.022947199642658234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,127,0.06773120164871216
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,127,0.049089598655700686
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,127,0.05568479895591736
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,127,0.03374719917774201
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,127,0.029097598791122437
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,127,0.02916640043258667
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,127,0.02924480140209198
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,127,0.06390399932861328
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,127,0.04136320054531097
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,127,0.043433600664138795
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,127,0.027193599939346315
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,127,0.02298080027103424
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,127,0.02301120012998581
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,127,0.022916799783706664
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,255,0.0839631974697113
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,255,0.06533920168876647
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,255,0.05956799983978271
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,255,0.04704320132732391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,255,0.043756800889968875
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,255,0.04337919950485229
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,255,0.04177600145339966
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,255,0.06808320283889771
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,255,0.050316798686981204
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,255,0.04554559886455536
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,255,0.02855679988861084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,255,0.023073600232601167
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,255,0.023600000143051147
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,255,0.02306399941444397
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,511,0.1073248028755188
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,511,0.09563040137290954
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,511,0.09675840139389039
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,511,0.07736799716949463
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,511,0.07251039743423462
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,511,0.07013599872589112
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,511,0.0717248022556305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,511,0.08110719919204712
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,511,0.06351360082626342
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,511,0.05966879725456238
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,511,0.044206398725509646
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,511,0.04101600050926209
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,511,0.03934240043163299
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,511,0.03914560079574585
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,1023,0.16113280057907103
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,1023,0.1484287977218628
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,1023,0.16212160587310792
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,1023,0.12861759662628175
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,1023,0.12346880435943604
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,1023,0.1214143991470337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,1023,0.12458720207214355
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,1023,0.10731199979782105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,1023,0.09020000100135803
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,1023,0.09144960045814514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,1023,0.0683456003665924
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,1023,0.06405919790267944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,1023,0.06284319758415222
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,1023,0.06207839846611023
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,2047,0.2638751983642578
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,2047,0.26201279163360597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,2047,0.2988719940185547
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,2047,0.23537759780883788
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,2047,0.2310192108154297
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,2047,0.2285856008529663
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,2047,0.2314176082611084
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,2047,0.16439679861068726
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,2047,0.14012000560760499
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,2047,0.15450079441070558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,2047,0.11563359498977661
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,2047,0.11064159870147705
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,2047,0.10889279842376709
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,2047,0.10788480043411255
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,512,1,1,4095,0.4792111873626709
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,512,1,2,4095,0.4803296089172363
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,512,1,4,4095,0.557535982131958
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,512,1,8,4095,0.4468111991882324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,512,1,16,4095,0.43921599388122556
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,512,1,32,4095,0.4365119934082031
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,512,1,64,4095,0.436187219619751
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,512,1,1,4095,0.2731647968292236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,512,1,2,4095,0.24209120273590087
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,512,1,4,4095,0.27866559028625487
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,512,1,8,4095,0.207588791847229
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,512,1,16,4095,0.20217759609222413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,512,1,64,4095,0.19936319589614868
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,512,1,32,4095,0.20054879188537597
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,1,0.1147663950920105
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,1,0.07440320253372193
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,1,0.09453759789466858
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,1,0.05176159739494324
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,1,0.04747680127620697
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,1,0.04549919962882996
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,1,0.04594559967517853
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,1,0.06816959977149964
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,1,0.11533919572830201
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,1,0.07612320184707641
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,1,0.04352959990501404
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,1,0.03722079992294312
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,1,0.03665120005607605
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,1,0.03507040143013
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,3,0.11476800441741944
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,3,0.07461599707603454
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,3,0.09481279850006104
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,3,0.05148000121116638
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,3,0.04742400050163269
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,3,0.045614400506019594
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,3,0.04551840126514435
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,3,0.11505600214004516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,3,0.06806399822235107
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,3,0.07619839906692505
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,3,0.04353919923305512
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,3,0.03722560107707977
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,3,0.03530240058898926
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,3,0.035359999537467955
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,7,0.11467839479446411
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,7,0.07465280294418335
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,7,0.09468479752540589
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,7,0.05156159996986389
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,7,0.04765279889106751
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,7,0.047224000096321106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,7,0.04622080028057098
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,7,0.11504000425338745
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,7,0.06808480024337768
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,7,0.07613919973373413
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,7,0.0435232013463974
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,7,0.037324801087379456
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,7,0.03539359867572785
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,7,0.03532640039920807
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,15,0.11530400514602661
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,15,0.07487679719924926
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,15,0.0947983980178833
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,15,0.05259199738502503
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,15,0.047495999932289125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,15,0.04755040109157562
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,15,0.046798399090766905
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,15,0.11512800455093383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,15,0.06827359795570373
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,15,0.07713599801063538
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,15,0.04359039962291718
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,15,0.03732799887657166
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,15,0.035504001379013064
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,15,0.03523840010166168
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,31,0.11512800455093383
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,31,0.07498400211334229
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,31,0.09463040232658386
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,31,0.053832000494003295
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,31,0.04763840138912201
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,31,0.047251200675964354
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,31,0.0458079993724823
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,31,0.11505759954452514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,31,0.07015680074691773
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,31,0.07786399722099305
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,31,0.04349919855594635
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,31,0.037303999066352844
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,31,0.0363072007894516
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,31,0.03547520041465759
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,63,0.11571999788284301
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,63,0.07701600193977357
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,63,0.09466239809989929
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,63,0.05589119791984558
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,63,0.04949280023574829
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,63,0.04753279983997345
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,63,0.0475488007068634
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,63,0.11507680416107177
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,63,0.07009599804878235
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,63,0.07866880297660828
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,63,0.0439087986946106
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,63,0.03731679916381836
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,63,0.03530719876289368
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,63,0.035412800312042234
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,127,0.12197279930114746
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,127,0.08584319949150085
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,127,0.09682719707489014
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,127,0.06414560079574586
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,127,0.05603359937667847
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,127,0.0546671986579895
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,127,0.05410879850387573
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,127,0.11516000032424926
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,127,0.07008960247039794
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,127,0.0782800018787384
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,127,0.04765760004520416
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,127,0.03939520120620728
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,127,0.03607999980449676
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,127,0.035488000512123107
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,255,0.15050400495529176
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,255,0.11325600147247314
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,255,0.10680320262908935
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,255,0.07872160077095032
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,255,0.07130879759788514
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,255,0.07017120122909545
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,255,0.06918879747390747
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,255,0.12105760574340821
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,255,0.0824400007724762
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,255,0.07798879742622375
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,255,0.051235198974609375
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,255,0.04385280013084412
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,255,0.0433135986328125
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,255,0.04187839925289154
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,511,0.19912960529327392
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,511,0.16275039911270142
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,511,0.17730239629745484
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,511,0.13346079587936402
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,511,0.12524319887161256
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,511,0.12068159580230713
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,511,0.12254400253295898
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,511,0.10486079454421997
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,511,0.1492784023284912
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,511,0.10554720163345337
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,511,0.07340160012245178
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,511,0.06587679982185364
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,511,0.06406239867210388
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,511,0.06235520243644714
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,1023,0.3059567928314209
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,1023,0.26431999206542967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,1023,0.3106031894683838
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,1023,0.22542879581451417
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,1023,0.2181936025619507
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,1023,0.21591360569000245
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,1023,0.2136415958404541
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,1023,0.20273280143737793
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,1023,0.1545807957649231
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,1023,0.16857919692993165
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,1023,0.11713919639587403
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,1023,0.11064800024032592
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,1023,0.10711840391159058
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,1023,0.10503519773483276
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,128,1024,1,1,2047,0.5123536109924316
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,64,1024,1,2,2047,0.4761807918548584
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,32,1024,1,4,2047,0.5787871837615967
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,16,1024,1,8,2047,0.431928014755249
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,8,1024,1,16,2047,0.4217807769775391
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,4,1024,1,32,2047,0.4175920009613037
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,float16,2,1024,1,64,2047,0.4146880149841309
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,128,1024,1,1,2047,0.3179487943649292
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,64,1024,1,2,2047,0.25030720233917236
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,32,1024,1,4,2047,0.2924000024795532
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,16,1024,1,8,2047,0.2054879903793335
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,8,1024,1,16,2047,0.19693119525909425
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,4,1024,1,32,2047,0.1929808020591736
SGLang,0.5.9,NVIDIA GB300,mla_generation,trtllm_mla,float16,fp8,2,1024,1,64,2047,0.19072959423065186
