Asymmetric Matrix Multiplication S1-S5 on H200
0.211Indicator S1PyTorch
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PyTorchPrecision=FP32, Reference=true2026.04 | 0.211 | 0.11 | 3.858 | 4.59 | 6.071 | 1.2 | |
| PyTorchPrecision=FP162026.04 | 0.128 | 0.121 | 3.144 | 3.161 | 3.183 | 0.866 | |
| SIMPLETESTF32=false2026.04 | 0.104 | 0.115 | 1.147 | 1.305 | 1.515 | 0.484 | |
| CUDA AgentPrecision=TF322026.04 | 0.082 | 0.075 | 3.338 | 3.29 | 3.424 | 0.747 | |
| SIMPLETESTF32=true2026.04 | 0.082 | 0.075 | — | — | — | 0.44 |