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 ERNIE-Doc Base (151M, WT103)?
Yes — here's how fast and up to what context, with real measured numbers.
ERNIE-Doc Base (151M, WT103) on the Instinct MI440X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 6037k | 40887.4-49973.5 t/s | ~83.4-102 (slow) |
| Q8 | yes | 6039k | 81774.8-99947 t/s | ~166.9-204 (slow) |
| Q4 | yes | 6040k | 146026.5-178476.8 t/s | ~298-364.2 (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 Instinct MI440X run ERNIE-Doc Base (151M, WT103)?▶
Yes. At Q4 it generates ~146026.5-178476.8 tok/s and fits up to 6040k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~298-364.2 tok/s.
How much context fits?▶
At Q4, up to about 6040k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.