INFERENTIA 2 vs RTX 4090
The INFERENTIA 2 has 32 GB of memory against 24 GB on the RTX 4090.
| Measure | AWS Inferentia2 | NVIDIA GeForce RTX 4090 |
|---|---|---|
| Lowest per GPU-hour | — | $0.16 |
| Median per GPU-hour | — | $0.58 |
| On-demand from | — | $0.19 |
| Spot from | — | $0.16 |
| Reserved from | — | $0.18 |
| Lowest per A100-equivalent hour | — | $0.25 |
| Live offers | 0 | 427 |
| Providers | 0 | 91 |
| Memory | 32 GB HBM2e | 24 GB GDDR6X |
| Memory bandwidth | 820 GB/s | 1,008 GB/s |
| FP16 / BF16 | 190 TFLOPS | 165.2 TFLOPS |
| FP8 | 190 TFLOPS | 330.3 TFLOPS |
| Performance vs A100 80GB | 0.52× | 0.63× |
| Architecture | Inferentia2 (NeuronCore v2) | Ada Lovelace |
| Power | — | 450 W |
Where the offers sit
No live offers right now.