A100 SXM 80GB vs INFERENTIA 2
The A100 SXM 80GB has 80 GB of memory against 32 GB on the INFERENTIA 2.
| Measure | NVIDIA A100 80GB SXM4 | AWS Inferentia2 |
|---|---|---|
| Lowest per GPU-hour | $0.61 | — |
| Median per GPU-hour | $1.71 | — |
| On-demand from | $0.68 | — |
| Spot from | $0.86 | — |
| Reserved from | $0.61 | — |
| Lowest per A100-equivalent hour | $0.61 | — |
| Live offers | 288 | 0 |
| Providers | 68 | 0 |
| Memory | 80 GB HBM2e | 32 GB HBM2e |
| Memory bandwidth | 2,039 GB/s | 820 GB/s |
| FP16 / BF16 | 312 TFLOPS | 190 TFLOPS |
| FP8 | — | 190 TFLOPS |
| Performance vs A100 80GB | 1× | 0.52× |
| Architecture | Ampere | Inferentia2 (NeuronCore v2) |
| Power | 400 W | — |
Where the offers sit
No live offers right now.