// RUN โ€” FITS

Can the Intel Gaudi 3 run CLIP (ViT L/14@336px)?

Yes โ€” here's how fast and up to what context, with real measured numbers.

CLIP (ViT L/14@336px) on the Intel Gaudi 3

QuantFits?Max contextSpeedBeyond context
FP16 yes 1763k 2900-3600 t/s ~31.4-38.9 (slow)
Q8 yes 1768k 5800-7200 t/s ~62.7-77.8 (slow)
Q4 yes 1771k 10357.1-12857.1 t/s ~112-139 (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 CLIP (ViT L/14@336px)?โ–ถ

Yes. At Q4 it generates ~10357.1-12857.1 tok/s and fits up to 1771k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~112-139 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1771k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.