A16 vs RTX 4090
The cheapest A16 ($0.058) costs 64% less per hour than the cheapest RTX 4090 ($0.16).
Per unit of performance, the RTX 4090 is 53% cheaper: $0.25 versus $0.53 per A100-equivalent hour.
The RTX 4090 has 24 GB of memory against 16 GB on the A16.
| Measure | A16 | NVIDIA GeForce RTX 4090 |
|---|---|---|
| Lowest per GPU-hour | $0.058 | $0.16 |
| Median per GPU-hour | $0.53 | $0.58 |
| On-demand from | $0.082 | $0.19 |
| Spot from | — | $0.16 |
| Reserved from | $0.058 | $0.18 |
| Lowest per A100-equivalent hour | $0.53 | $0.25 |
| Live offers | 26 | 427 |
| Providers | 3 | 91 |
| Memory | 16 GB GDDR6 | 24 GB GDDR6X |
| Memory bandwidth | 200 GB/s | 1,008 GB/s |
| FP16 / BF16 | 35.9 TFLOPS | 165.2 TFLOPS |
| FP8 | — | 330.3 TFLOPS |
| Performance vs A100 80GB | 0.11× | 0.63× |
| Architecture | Ampere | Ada Lovelace |
| Power | 62 W | 450 W |