AMD Instinct MI300X 192GB vs NVIDIA B300 SXM
Instinct MI300X 192GB has 33% less memory than B300 SXM (192 GB vs 288 GB). Its memory bandwidth is 34% less (5.3 vs 8 TB/s). In the cloud, the median on-demand Instinct MI300X 192GB price is $2.99 per GPU-hour against $7.64 for the B300 SXM. Per gigabyte of GPU memory, that is $0.016 vs $0.027 per hour.
Specifications
Manufacturer figures as recorded by Kovara. Higher memory and bandwidth are highlighted; neither is a benchmark of your workload.
| Instinct MI300X 192GB | B300 SXM | |
|---|---|---|
| Manufacturer | AMD | NVIDIA |
| Architecture | CDNA 3 | Blackwell Ultra |
| Memory | 192 GB HBM3 | 288 GB HBM3E |
| Memory bandwidth | 5.3 TB/s | 8 TB/s |
| Power (TDP) | 750 W | 1400 W |
| Form factor | OAM Module | SXM |
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.
| Instinct MI300X 192GB | B300 SXM | |
|---|---|---|
| Lowest price | $2.59 | $6.60 |
| Median price | $2.99 | $7.64 |
| Highest price | $3.45 | $12.66 |
| Median per GB of memory | $0.016 | $0.027 |
| Public offers (all terms) | 8 | 30 |
| Providers | 3 | 12 |
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 Instinct MI300X 192GB or the B300 SXM?
Right now the cheapest on-demand Instinct MI300X 192GB Kovara tracks is $2.59 per GPU-hour and the cheapest B300 SXM is $6.60. Medians are $2.99 and $7.64. 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 B300 SXM has more memory: 288 GB against 192 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.