List your company
ResearchFormerly WizWand

Matrix Multiplication Performance on RTX 4090 (Sizes 2048-8192)

0.1002Latency (ms)RT-Lynx
-0.3183362.5067825.33198.157018May 26, 2026
Updated 5mo ago

Evaluation results

MethodLinks
RT-LynxMatrix Shape (M=N)=2048, Matrix Shape (K)=30722026.050.10021.6110.97
GEMMMatrix Shape (M=N)=2048, Matrix Shape (K)=30722026.050.1618——
cuSparseLtMatrix Shape (M=N)=2048, Matrix Shape (K)=30722026.050.24080.6764.04
RT-LynxMatrix Shape (M=N)=4096, Matrix Shape (K)=30722026.050.3541.785.79
RT-LynxMatrix Shape (M=N)=2048, Matrix Shape (K)=122882026.050.40031.6716.64
Pytorch-SpMMMatrix Shape (M=N)=2048, Matrix Shape (K)=30722026.050.52260.3138.18
cuSparseLtMatrix Shape (M=N)=4096, Matrix Shape (K)=30722026.050.59561.0647.61
Pytorch-SpMMMatrix Shape (M=N)=4096, Matrix Shape (K)=30722026.050.60531.0444.31
GEMMMatrix Shape (M=N)=4096, Matrix Shape (K)=30722026.050.6301——
GEMMMatrix Shape (M=N)=2048, Matrix Shape (K)=122882026.050.667——
Pytorch-SpMMMatrix Shape (M=N)=2048, Matrix Shape (K)=122882026.050.8850.7563.24
cuSparseLtMatrix Shape (M=N)=2048, Matrix Shape (K)=122882026.050.91220.7366.72
RT-LynxMatrix Shape (M=N)=8192, Matrix Shape (K)=30722026.051.44661.794.38
RT-LynxMatrix Shape (M=N)=4096, Matrix Shape (K)=122882026.051.51981.7313.09
Pytorch-SpMMMatrix Shape (M=N)=8192, Matrix Shape (K)=30722026.051.87291.3829.43
cuSparseLtMatrix Shape (M=N)=8192, Matrix Shape (K)=30722026.051.98691.330.51
Pytorch-SpMMMatrix Shape (M=N)=4096, Matrix Shape (K)=122882026.052.37811.1147.85
cuSparseLtMatrix Shape (M=N)=4096, Matrix Shape (K)=122882026.052.58271.0252.84
GEMMMatrix Shape (M=N)=8192, Matrix Shape (K)=30722026.052.585——
GEMMMatrix Shape (M=N)=4096, Matrix Shape (K)=122882026.052.6281——
RT-LynxMatrix Shape (M=N)=8192, Matrix Shape (K)=122882026.056.261.696.41
Pytorch-SpMMMatrix Shape (M=N)=8192, Matrix Shape (K)=122882026.057.47511.4129.8
cuSparseLtMatrix Shape (M=N)=8192, Matrix Shape (K)=122882026.057.66541.3835.18
GEMMMatrix Shape (M=N)=8192, Matrix Shape (K)=122882026.0510.5636——