// RUN โ€” FITS

Can the NVIDIA H200 run BIDAF?

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

BIDAF on the NVIDIA H200

QuantFits?Max contextSpeedBeyond context
FP16 yes 1956k 673846.2-803076.9 t/s ~5615.4-6692.3 (slow)
Q8 yes 1956k 1347692.3-1606153.8 t/s ~11230.8-13384.6 (slow)
Q4 yes 1956k 2406593.4-2868131.9 t/s ~20054.9-23901.1 (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 H200 run BIDAF?โ–ถ

Yes. At Q4 it generates ~2406593.4-2868131.9 tok/s and fits up to 1956k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~20054.9-23901.1 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1956k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.