New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Can the NVIDIA RTX 6000 Ada 48GB run Octo-Base?
Yes — here's how fast and up to what context, with real measured numbers.
Octo-Base on the NVIDIA RTX 6000 Ada 48GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 648k | 3767.7-4490.3 t/s | ~157-187.1 (slow) |
| Q8 | yes | 650k | 7535.5-8980.6 t/s | ~314-374.2 (slow) |
| Q4 | yes | 650k | 13456.2-16036.9 t/s | ~560.7-668.2 (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 RTX 6000 Ada 48GB run Octo-Base?▶
Yes. At Q4 it generates ~13456.2-16036.9 tok/s and fits up to 650k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~560.7-668.2 tok/s.
How much context fits?▶
At Q4, up to about 650k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.