New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Can the NVIDIA H200 run FvFold?
Yes — here's how fast and up to what context, with real measured numbers.
FvFold on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1956k | 190232.4-226715.3 t/s | ~1585.3-1889.3 (slow) |
| Q8 | yes | 1956k | 380464.8-453430.6 t/s | ~3170.5-3778.6 (slow) |
| Q4 | yes | 1956k | 679401.4-809697.6 t/s | ~5661.7-6747.5 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.
FAQ
Can the NVIDIA H200 run FvFold?▶
Yes. At Q4 it generates ~679401.4-809697.6 tok/s and fits up to 1956k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~5661.7-6747.5 tok/s.
How much context fits?▶
At Q4, up to about 1956k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.