…
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →| Spec / price | RTX PRO 6000 Blackwell | L40S |
|---|---|---|
| Memory | 96 GB GDDR7better | 48 GB GDDR6 |
| Memory bandwidth | 1,792 GB/sbetter | 864 GB/s |
| FP16 / BF16 (dense) | 0.504 PFbetter | 0.362 PF |
| FP8 (dense) | 1.01 PFbetter | 0.733 PF |
| FP4 (dense) | 2.02 PF | — |
| INT8 (dense) | 1.01 PFbetter | 0.733 PF |
| TF32 (dense) | 0.126 PF | 0.183 PFbetter |
| FP32 (dense) | 0.126 PFbetter | 0.0916 PF |
| FP64 (dense) | 0.002 PFbetter | 0.0014 PF |
| Power | 600 W | 350 W |
| Architecture | Blackwell (2025) | Ada Lovelace (2023) |
| Lowest on-demand | $0.59better | $0.60 |
| Median on-demand | $1.56 | $1.50better |
| Lowest spot | $0.32better | $0.38 |
| FP16/BF16 $ per PFLOP·h | $1.17better | $1.66 |
| Providers | 17 | 17 |
The RTX PRO 6000 Blackwell has 1.4× more dense FP16/BF16 compute than the L40S. The RTX PRO 6000 Blackwell has 2.1× more memory bandwidth than the L40S. Real-world speedups depend on the workload: training and prefill scale with compute, while LLM decoding scales mostly with memory bandwidth.
Right now the RTX PRO 6000 Blackwell is cheaper per unit of compute: $1.17 vs $1.66 per PFLOP-hour at the lowest on-demand prices.
Running one GPU around the clock at the lowest on-demand price costs about $431 per month for the RTX PRO 6000 Blackwell and $440 for the L40S.
The RTX PRO 6000 Blackwell has more memory: 96 GB vs 48 GB.