Kovara
Loading your page…
You can continue using the navigation while this page loads.
You can continue using the navigation while this page loads.
H100 SXM has 67% more memory than L40S 48GB (80 GB vs 48 GB). Its memory bandwidth is 288% more (3.35 vs 0.864 TB/s). In the cloud, the median on-demand H100 SXM price is $3.85 per GPU-hour against $1.50 for the L40S 48GB. Per gigabyte of GPU memory, that is $0.048 vs $0.031 per hour.
Manufacturer figures as recorded by Kovara. Higher memory and bandwidth are highlighted; neither is a benchmark of your workload.
| H100 SXM | L40S 48GB | |
|---|---|---|
| Manufacturer | NVIDIA | NVIDIA |
| Architecture | Hopper | Ada Lovelace |
| Memory | 80 GB HBM3 | 48 GB GDDR6 |
| Memory bandwidth | 3.35 TB/s | 0.864 TB/s |
| Power (TDP) | 700 W | 350 W |
| Form factor | SXM | PCIe dual-slot |
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.
| H100 SXM | L40S 48GB | |
|---|---|---|
| Lowest price | $1.99 | $0.97 |
| Median price | $3.85 | $1.50 |
| Highest price | $12.29 | $2.25 |
| Median per GB of memory | $0.048 | $0.031 |
| Public offers (all terms) | 59 | 38 |
| Providers | 21 | 9 |
Lowest tracked price per provider. Open a provider to see every offer, region and term.
Right now the cheapest on-demand H100 SXM Kovara tracks is $1.99 per GPU-hour and the cheapest L40S 48GB is $0.97. Medians are $3.85 and $1.50. The cheaper GPU per hour is not always cheaper per job: a faster or larger GPU can finish sooner or need fewer units.
The H100 SXM has more memory: 80 GB against 48 GB. Memory decides which models and batch sizes fit on one GPU.
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.