NVIDIA GH200 96GB HBM3 vs NVIDIA H100 NVL 94GB
GH200 96GB HBM3 has about the same memory than H100 NVL 94GB (96 GB vs 94 GB). Its memory bandwidth is about the same (4 vs 3.9 TB/s). In the cloud, the median on-demand GH200 96GB HBM3 price is $5.19 per GPU-hour against $8.66 for the H100 NVL 94GB. Per gigabyte of GPU memory, that is $0.054 vs $0.092 per hour.
Specifications
Manufacturer figures as recorded by Kovara. Higher memory and bandwidth are highlighted; neither is a benchmark of your workload.
| GH200 96GB HBM3 | H100 NVL 94GB | |
|---|---|---|
| Manufacturer | NVIDIA | NVIDIA |
| Architecture | Hopper + Grace | Hopper |
| Memory | 96 GB HBM3 | 94 GB HBM3 |
| Memory bandwidth | 4 TB/s | 3.9 TB/s |
| Power (TDP) | 900 W | 400 W |
| Form factor | Grace Hopper Superchip | PCIe |
Cloud prices today
Public offers Kovara tracks, normalised to US dollars per GPU-hour. Prices use on-demand terms where available; spot and reserved offers are on each GPU’s price pages. Lower is highlighted.
| GH200 96GB HBM3 | H100 NVL 94GB | |
|---|---|---|
| Lowest price | $2.29 | $2.92 |
| Median price | $5.19 | $8.66 |
| Highest price | $6.50 | $10.12 |
| Median per GB of memory | $0.054 | $0.092 |
| Public offers (all terms) | 5 | 60 |
| Providers | 3 | 3 |
Where to rent each
Lowest tracked price per provider. Open a provider to see every offer, region and term.
Questions
Which is cheaper to rent, the GH200 96GB HBM3 or the H100 NVL 94GB?
Right now the cheapest on-demand GH200 96GB HBM3 Kovara tracks is $2.29 per GPU-hour and the cheapest H100 NVL 94GB is $2.92. Medians are $5.19 and $8.66. The cheaper GPU per hour is not always cheaper per job: a faster or larger GPU can finish sooner or need fewer units.
Which has more memory?
The GH200 96GB HBM3 has more memory: 96 GB against 94 GB. Memory decides which models and batch sizes fit on one GPU.
Are these benchmark results?
No. This page compares manufacturer specifications and public prices. Real performance depends on the model, precision, software and interconnect. Kova can help you estimate cost for a specific workload.