| Precision | Peak (TFLOPS) | Per GPU (TFLOPS) | TFLOPS/W |
|---|---|---|---|
| FP64 | 78 | 9.7 | 12.93 |
| FP32 | 156 | 20 | 26.00 |
| TF32 | 1,248 | 156 | 208.00 |
| FP16 | 2,496 | 312 | 416.00 |
| BF16 | 2,496 | 312 | 416.00 |
| INT8 | 4,992 | 624 | 832.00 |
All figures are dense. Vendors commonly headline the sparse number, which is twice the dense one.
NVSwitch + NVLink
DGX A100 with 320GB total GPU memory