| Precision | Peak (TFLOPS) | Per GPU (TFLOPS) | TFLOPS/W |
|---|---|---|---|
| FP64 | 39 | 9.7 | 12.12 |
| FP32 | 78 | 20 | 24.38 |
| TF32 | 624 | 156 | 195.00 |
| FP16 | 1,248 | 312 | 390.00 |
| BF16 | 1,248 | 312 | 390.00 |
| INT8 | 2,496 | 624 | 780.00 |
All figures are dense. Vendors commonly headline the sparse number, which is twice the dense one.
NVLink
HGX A100 4-GPU configuration