// RUN โ€” FITS

Can the Intel Gaudi 3 run CARP?

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

CARP on the Intel Gaudi 3

QuantFits?Max contextSpeedBeyond context
FP16 yes 1754k 1668.7-2071.5 t/s ~18-22.4 (slow)
Q8 yes 1764k 3337.5-4143.1 t/s ~36.1-44.8 (slow)
Q4 yes 1768k 5959.8-7398.4 t/s ~64.4-80 (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 CARP?โ–ถ

Yes. At Q4 it generates ~5959.8-7398.4 tok/s and fits up to 1768k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~64.4-80 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1768k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.