A40 vs RTX 5090
The cheapest RTX 5090 ($0.22) costs 6% less per hour than the cheapest A40 ($0.23).
Per unit of performance, the RTX 5090 is 57% cheaper: $0.24 versus $0.55 per A100-equivalent hour.
The A40 has 48 GB of memory against 32 GB on the RTX 5090.
| Measure | NVIDIA A40 48GB | NVIDIA GeForce RTX 5090 |
|---|---|---|
| Lowest per GPU-hour | $0.23 | $0.22 |
| Median per GPU-hour | $0.79 | $0.87 |
| On-demand from | $0.30 | $0.34 |
| Spot from | — | $0.25 |
| Reserved from | $0.23 | $0.22 |
| Lowest per A100-equivalent hour | $0.55 | $0.24 |
| Live offers | 72 | 405 |
| Providers | 29 | 84 |
| Memory | 48 GB GDDR6 | 32 GB GDDR7 |
| Memory bandwidth | 696 GB/s | 1,792 GB/s |
| FP16 / BF16 | 149.7 TFLOPS | 209.5 TFLOPS |
| FP8 | — | 419 TFLOPS |
| Performance vs A100 80GB | 0.42× | 0.92× |
| Architecture | Ampere | Blackwell |
| Power | 300 W | 575 W |