🔌
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 Qwen-VLA?
Yes — here's how fast and up to what context, with real measured numbers.
Qwen-VLA on the Instinct MI455X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 2942k | 1198.8-1465.2 t/s | ~2.4-3 (slow) |
| Q8 | yes | 2981k | 2397.7-2930.5 t/s | ~4.9-6 (slow) |
| Q4 | yes | 2999k | 4281.6-5233 t/s | ~8.7-10.7 (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 Qwen-VLA?▶
Yes. At Q4 it generates ~4281.6-5233 tok/s and fits up to 2999k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~8.7-10.7 tok/s.
How much context fits?▶
At Q4, up to about 2999k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.