Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
M3 Ultra (192 GB) has 16 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, DGX Spark (GB10) delivers 4.6× higher FP16 throughput.
Specifications
| DGX Spark (GB10) | M3 Ultra (192 GB) | |
|---|---|---|
| VRAM | 128 GB | 144 GB |
| VRAM Type | LPDDR5X | LPDDR5 |
| Memory Bandwidth | 0.3 TB/s | 0.8 TB/s |
| FP16 Performance | 125 TFLOPS | 27 TFLOPS |
| Manufacturer | NVIDIA | Apple |
| FP8 Support | Yes | No |
| FP4 Support | Yes | No |
Price / Performance
Based on cheapest single-GPU on-demand pricing. Lower $/TFLOP = better compute value; lower $/GB = better memory value.
| DGX Spark (GB10) | M3 Ultra (192 GB) | |
|---|---|---|
| $/hr (cheapest) | — | — |
| $/TFLOP (compute value) | — | — |
| $/GB VRAM (memory value) | — | — |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
DGX Spark (GB10)
No cloud pricing available.
M3 Ultra (192 GB)
No cloud pricing available.
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
DGX Spark (GB10) (1000 models)
M3 Ultra (192 GB) (1000 models)
You might also compare…
Pricing data refreshed hourly · Last updated June 1, 2026 · Browse all comparisons