Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
MI300X has 96 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, MI300X delivers 1.3× higher FP16 throughput. At $1.99/hr vs $2.29/hr, MI300X is the more cost-efficient choice for inference. MI300X supports a broader range of models (659 vs 637 from this catalog), giving more flexibility.
Specifications
| GH200 | MI300X | |
|---|---|---|
| VRAM | 96 GB | 192 GB |
| VRAM Type | HBM3 | HBM3 |
| Memory Bandwidth | 4.0 TB/s | 5.3 TB/s |
| FP16 Performance | 990 TFLOPS | 1307 TFLOPS |
| Manufacturer | NVIDIA | AMD |
| FP8 Support | Yes | Yes |
| FP4 Support | No | No |
Price / Performance
Based on cheapest single-GPU on-demand pricing. Lower $/TFLOP = better compute value; lower $/GB = better memory value.
| GH200 | MI300X | |
|---|---|---|
| $/hr (cheapest) | $2.29 | $1.99✓ best |
| $/TFLOP (compute value) | $0.0023 | $0.0015✓ best |
| $/GB VRAM (memory value) | $0.0239 | $0.0104✓ best |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
GH200 (637 models)
MI300X (659 models)
You might also compare…
Pricing data refreshed hourly · Last updated April 11, 2026 · Browse all comparisons