New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Can the Intel Gaudi 3 run base LM+GNN+kNN?
Yes — here's how fast and up to what context, with real measured numbers.
base LM+GNN+kNN on the Intel Gaudi 3
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1766k | 3916.1-4861.3 t/s | ~42.3-52.6 (slow) |
| Q8 | yes | 1770k | 7832.1-9722.6 t/s | ~84.7-105.1 (slow) |
| Q4 | yes | 1772k | 13985.9-17361.8 t/s | ~151.2-187.7 (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 Intel Gaudi 3 run base LM+GNN+kNN?▶
Yes. At Q4 it generates ~13985.9-17361.8 tok/s and fits up to 1772k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~151.2-187.7 tok/s.
How much context fits?▶
At Q4, up to about 1772k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.