H100 SXM 80GB vs INFERENTIA 2
The H100 SXM 80GB has 80 GB of memory against 32 GB on the INFERENTIA 2.
| Measure | NVIDIA H100 80GB SXM5 | AWS Inferentia2 |
|---|---|---|
| Lowest per GPU-hour | $0.50 | — |
| Median per GPU-hour | $3.51 | — |
| On-demand from | $0.58 | — |
| Spot from | $0.50 | — |
| Reserved from | $1.41 | — |
| Lowest per A100-equivalent hour | $0.17 | — |
| Live offers | 317 | 0 |
| Providers | 83 | 0 |
| Memory | 80 GB HBM3 | 32 GB HBM2e |
| Memory bandwidth | 3,350 GB/s | 820 GB/s |
| FP16 / BF16 | 989 TFLOPS | 190 TFLOPS |
| FP8 | 1,979 TFLOPS | 190 TFLOPS |
| Performance vs A100 80GB | 3× | 0.52× |
| Architecture | Hopper | Inferentia2 (NeuronCore v2) |
| Power | 700 W | — |
Where the offers sit
No live offers right now.