// RUN โ€” FITS

Can the NVIDIA L40S 48GB run CodeT5-large?

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

CodeT5-large on the NVIDIA L40S 48GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 627k 409.6-488.1 t/s ~19-22.6 (slow)
Q8 yes 639k 819.1-976.2 t/s ~37.9-45.2 (slow)
Q4 yes 644k 1462.7-1743.2 t/s ~67.7-80.7 (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 NVIDIA L40S 48GB run CodeT5-large?โ–ถ

Yes. At Q4 it generates ~1462.7-1743.2 tok/s and fits up to 644k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~67.7-80.7 tok/s.

How much context fits?โ–ถ

At Q4, up to about 644k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.