…
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →| Spec / price | H200 SXM | B200 SXM |
|---|---|---|
| Memory | 141 GB HBM3e | 180 GB HBM3ebetter |
| Memory bandwidth | 4,800 GB/s | 7,700 GB/sbetter |
| FP16 / BF16 (dense) | 0.989 PF | 2.25 PFbetter |
| FP8 (dense) | 1.98 PF | 4.5 PFbetter |
| FP4 (dense) | — | 9 PF |
| INT8 (dense) | 1.98 PF | 4.5 PFbetter |
| TF32 (dense) | 0.495 PF | 1.1 PFbetter |
| FP32 (dense) | 0.067 PF | 0.075 PFbetter |
| FP64 (dense) | 0.067 PFbetter | 0.037 PF |
| Power | 700 W | 1000 W |
| Architecture | Hopper (2024) | Blackwell (2024) |
| Lowest on-demand | $2.10better | $3.74 |
| Median on-demand | $3.99better | $6.80 |
| Lowest spot | $0.79better | $0.99 |
| FP16/BF16 $ per PFLOP·h | $2.12 | $1.66better |
| Providers | 22 | 15 |
The B200 SXM has 2.3× more dense FP16/BF16 compute than the H200 SXM. The B200 SXM has 1.6× more memory bandwidth than the H200 SXM. Real-world speedups depend on the workload: training and prefill scale with compute, while LLM decoding scales mostly with memory bandwidth.
Right now the B200 SXM is cheaper per unit of compute: $1.66 vs $2.12 per PFLOP-hour at the lowest on-demand prices.
Running one GPU around the clock at the lowest on-demand price costs about $1,533 per month for the H200 SXM and $2,730 for the B200 SXM.
The B200 SXM has more memory: 180 GB vs 141 GB.