// RUN โ€” FITS

Can the AMD Instinct MI300X 192GB run Falcon-40B?

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

Falcon-40B on the AMD Instinct MI300X 192GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 290k 41.7-51 t/s ~0.3-0.4 (slow)
Q8 yes 412k 83.5-102 t/s ~0.6-0.8 (slow)
Q4 yes 466k 149.1-182.2 t/s ~1.1-1.4 (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 AMD Instinct MI300X 192GB run Falcon-40B?โ–ถ

Yes. At Q4 it generates ~149.1-182.2 tok/s and fits up to 466k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~1.1-1.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 466k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.