New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Can the Instinct MI440X run MT-DNN?
Yes — here's how fast and up to what context, with real measured numbers.
MT-DNN on the Instinct MI440X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 6031k | 18709.1-22866.7 t/s | ~38.2-46.7 (slow) |
| Q8 | yes | 6037k | 37418.2-45733.3 t/s | ~76.4-93.3 (slow) |
| Q4 | yes | 6039k | 66818.2-81666.7 t/s | ~136.4-166.7 (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 MI440X run MT-DNN?▶
Yes. At Q4 it generates ~66818.2-81666.7 tok/s and fits up to 6039k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~136.4-166.7 tok/s.
How much context fits?▶
At Q4, up to about 6039k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.