L20 vs RTX 4090
The cheapest RTX 4090 ($0.16) costs 55% less per hour than the cheapest L20 ($0.36).
Per unit of performance, the RTX 4090 is 65% cheaper: $0.25 versus $0.73 per A100-equivalent hour.
The L20 has 48 GB of memory against 24 GB on the RTX 4090.
| Measure | NVIDIA L20 48GB | NVIDIA GeForce RTX 4090 |
|---|---|---|
| Lowest per GPU-hour | $0.36 | $0.16 |
| Median per GPU-hour | $1.85 | $0.58 |
| On-demand from | $0.60 | $0.19 |
| Spot from | $0.36 | $0.16 |
| Reserved from | $0.48 | $0.18 |
| Lowest per A100-equivalent hour | $0.73 | $0.25 |
| Live offers | 136 | 427 |
| Providers | 5 | 91 |
| Memory | 48 GB GDDR6 | 24 GB GDDR6X |
| Memory bandwidth | 864 GB/s | 1,008 GB/s |
| FP16 / BF16 | 119.5 TFLOPS | 165.2 TFLOPS |
| FP8 | 239 TFLOPS | 330.3 TFLOPS |
| Performance vs A100 80GB | 0.49× | 0.63× |
| Architecture | Ada Lovelace | Ada Lovelace |
| Power | 275 W | 450 W |