NVIDIA A30 24GB vs NVIDIA GH200 96GB HBM3
A30 24GB has 75% less memory than GH200 96GB HBM3 (24 GB vs 96 GB). Its memory bandwidth is 77% less (0.933 vs 4 TB/s). In the cloud, the median on-demand A30 24GB price is $0.41 per GPU-hour against $5.19 for the GH200 96GB HBM3. Per gigabyte of GPU memory, that is $0.017 vs $0.054 per hour.
Specifications
Manufacturer figures as recorded by Kovara. Higher memory and bandwidth are highlighted; neither is a benchmark of your workload.
| A30 24GB | GH200 96GB HBM3 | |
|---|---|---|
| Manufacturer | NVIDIA | NVIDIA |
| Architecture | Ampere | Hopper + Grace |
| Memory | 24 GB HBM2 | 96 GB HBM3 |
| Memory bandwidth | 0.933 TB/s | 4 TB/s |
| Power (TDP) | 165 W | 900 W |
| Form factor | PCIe dual-slot FHFL | Grace Hopper Superchip |
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.
| A30 24GB | GH200 96GB HBM3 | |
|---|---|---|
| Lowest price | $0.35 | $2.29 |
| Median price | $0.41 | $5.19 |
| Highest price | $1.35 | $6.50 |
| Median per GB of memory | $0.017 | $0.054 |
| Public offers (all terms) | 9 | 5 |
| 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 A30 24GB or the GH200 96GB HBM3?
Right now the cheapest on-demand A30 24GB Kovara tracks is $0.35 per GPU-hour and the cheapest GH200 96GB HBM3 is $2.29. Medians are $0.41 and $5.19. 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 24 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.