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 TAPE Transformer?
Yes — here's how fast and up to what context, with real measured numbers.
TAPE Transformer on the Intel Gaudi 3
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1773k | 28236.8-35052.6 t/s | ~305.3-378.9 (slow) |
| Q8 | yes | 1773k | 56473.7-70105.3 t/s | ~610.5-757.9 (slow) |
| Q4 | yes | 1774k | 100845.9-125188 t/s | ~1090.2-1353.4 (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 TAPE Transformer?▶
Yes. At Q4 it generates ~100845.9-125188 tok/s and fits up to 1774k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~1090.2-1353.4 tok/s.
How much context fits?▶
At Q4, up to about 1774k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.