Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
M3 Ultra (192 GB) has 48 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, GH200 delivers 36.4× higher FP16 throughput. M3 Ultra (192 GB) supports a broader range of models (657 vs 637 from this catalog), giving more flexibility.
Specifications
| GH200 | M3 Ultra (192 GB) | |
|---|---|---|
| VRAM | 96 GB | 144 GB |
| VRAM Type | HBM3 | LPDDR5 |
| Memory Bandwidth | 4.0 TB/s | 0.8 TB/s |
| FP16 Performance | 990 TFLOPS | 27 TFLOPS |
| Manufacturer | NVIDIA | Apple |
| FP8 Support | Yes | No |
| 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 | M3 Ultra (192 GB) | |
|---|---|---|
| $/hr (cheapest) | $2.29 | — |
| $/TFLOP (compute value) | $0.0023 | — |
| $/GB VRAM (memory value) | $0.0239 | — |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
GH200
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Lambda | $2.29/hr | — |
M3 Ultra (192 GB)
No cloud pricing available.
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
GH200 (637 models)
M3 Ultra (192 GB) (657 models)
You might also compare…
Pricing data refreshed hourly · Last updated April 11, 2026 · Browse all comparisons