// RUN โ€” FITS

Can the NVIDIA L40S 48GB 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 NVIDIA L40S 48GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 640k 852.3-1015.8 t/s ~39.5-47 (slow)
Q8 yes 645k 1704.6-2031.6 t/s ~78.9-94.1 (slow)
Q4 yes 648k 3044-3627.8 t/s ~140.9-168 (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 CLIP (ViT L/14@336px)?โ–ถ

Yes. At Q4 it generates ~3044-3627.8 tok/s and fits up to 648k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~140.9-168 tok/s.

How much context fits?โ–ถ

At Q4, up to about 648k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.