🔌
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 AMD Instinct MI325X run Llama 3.2 90B?
Yes — here's how fast and up to what context, with real measured numbers.
Llama 3.2 90B on the AMD Instinct MI325X
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 173k | 21.3-26.1 t/s | ~0.1-0.2 (slow) |
| Q8 | yes | 444k | 42.7-52.1 t/s | ~0.3-0.3 (slow) |
| Q4 | yes | 563k | 76.2-93.1 t/s | ~0.5-0.6 (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 AMD Instinct MI325X run Llama 3.2 90B?▶
Yes. At Q4 it generates ~76.2-93.1 tok/s and fits up to 563k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~0.5-0.6 tok/s.
How much context fits?▶
At Q4, up to about 563k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.