🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the Instinct MI455X run InternVL2-Llama3-76B?
Yes — here's how fast and up to what context, with real measured numbers.
InternVL2-Llama3-76B on the Instinct MI455X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 744k | 81.2-99.3 t/s | ~0.2-0.2 (slow) |
| Q8 | yes | 976k | 162.5-198.6 t/s | ~0.3-0.4 (slow) |
| Q4 | yes | 1078k | 290.1-354.6 t/s | ~0.6-0.7 (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 Instinct MI455X run InternVL2-Llama3-76B?▶
Yes. At Q4 it generates ~290.1-354.6 tok/s and fits up to 1078k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~0.6-0.7 tok/s.
How much context fits?▶
At Q4, up to about 1078k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.