🔌
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 MI300X 192GB run Very Deep VAEs (ImageNet-64)?
Yes — here's how fast and up to what context, with real measured numbers.
Very Deep VAEs (ImageNet-64) on the AMD Instinct MI300X 192GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 2669k | 13356-16324 t/s | ~100.8-123.2 (slow) |
| Q8 | yes | 2671k | 26712-32648 t/s | ~201.6-246.4 (slow) |
| Q4 | yes | 2671k | 47700-58300 t/s | ~360-440 (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 MI300X 192GB run Very Deep VAEs (ImageNet-64)?▶
Yes. At Q4 it generates ~47700-58300 tok/s and fits up to 2671k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~360-440 tok/s.
How much context fits?▶
At Q4, up to about 2671k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.