| Precision | Peak (PFLOPS) | Per GPU (TFLOPS) | TFLOPS/W |
|---|---|---|---|
| FP64 | 0.1 | 34.0 | 32.38 |
| FP32 | 0.3 | 67.0 | 63.81 |
| TF32 | 2.0 | 494.5 | 941.90 |
| FP16 | 4.0 | 989.5 | 1884.76 |
| BF16 | 4.0 | 989.5 | 1884.76 |
| FP8 | 7.9 | 1979.0 | 3769.52 |
| INT8 | 7.9 | 1979.0 | 3769.52 |
All figures are dense. Vendors commonly headline the sparse number, which is twice the dense one.
NVLink
HGX H100 4-GPU configuration for smaller deployments