GPU comparison
MI300X vs H100
MI300X has 2.4× the memory of an H100 (192 vs 80 GB), more bandwidth and more dense FP8, for a lower monthly price. H100 wins when you depend on CUDA-only libraries or tuned Hopper recipes; for serving large models on ROCm, MI300X is the cost-efficient choice.
- Updated Nov 4, 2026
- Specs: vendor datasheets, dense Tensor figures
- Prices: live CoreGPU catalog
| Spec | AMD Instinct MI300X | NVIDIA H100 SXM |
|---|---|---|
| Architecture | CDNA 3 | Hopper |
| GPU memory | 192 GB HBM3 | 80 GB HBM3 |
| Memory bandwidth | 5.3 TB/s | 3.35 TB/s |
| FP8 dense | 2.6 PF | 1.98 PF |
| FP4 dense | n/a | n/a |
| Interconnect | Infinity Fabric · 896 GB/s | NVLink 4 · 900 GB/s |
| Power (TDP) | 750 W | 700 W |
| Price per GPU / month | $1,749 | $1,999 |
| Price per GB of memory | $9.11 | $24.99 |
| Price per FP8 PFLOPS | $673 | $1,010 |
| Availability | 41 available | Only 3 left |
Green marks the better value on each line. Prices are per GPU per month at CoreGPU, at least 25% below the market median.
- Serving 70B–405B models with fewer GPUs
- Memory-bound decoding
- The lowest monthly price for an HBM GPU
$1,749 /GPU/moConfigure MI300X
- CUDA-only kernels and libraries
- Recipes tuned for Hopper
- Mixed training and development work
$1,999 /GPU/moConfigure H100