// RUN โ€” FITS

Can the Intel Arc B580 12GB run SciBERT?

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

SciBERT on the Intel Arc B580 12GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 142k 1202.2-1492.4 t/s ~105.5-130.9 (slow)
Q8 yes 144k 2404.4-2984.7 t/s ~210.9-261.8 (slow)
Q4 yes 145k 4293.5-5329.9 t/s ~376.6-467.5 (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 Arc B580 12GB run SciBERT?โ–ถ

Yes. At Q4 it generates ~4293.5-5329.9 tok/s and fits up to 145k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~376.6-467.5 tok/s.

How much context fits?โ–ถ

At Q4, up to about 145k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.