// RUN โ€” FITS

Can the Instinct MI440X run MegaMolBART?

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

MegaMolBART on the Instinct MI440X

QuantFits?Max contextSpeedBeyond context
FP16 yes 6040k 137200-167688.9 t/s ~280-342.2 (slow)
Q8 yes 6041k 274400-335377.8 t/s ~560-684.4 (slow)
Q4 yes 6041k 490000-598888.9 t/s ~1000-1222.2 (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 Instinct MI440X run MegaMolBART?โ–ถ

Yes. At Q4 it generates ~490000-598888.9 tok/s and fits up to 6041k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~1000-1222.2 tok/s.

How much context fits?โ–ถ

At Q4, up to about 6041k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.