// RUN โ€” FITS

Can the NVIDIA DGX Spark run DistilProtBert?

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

DistilProtBert on the NVIDIA DGX Spark

QuantFits?Max contextSpeedBeyond context
FP16 yes 1728k 433.2-516.3 t/s ~15.9-18.9 (slow)
Q8 yes 1731k 866.5-1032.7 t/s ~31.7-37.8 (slow)
Q4 yes 1733k 1547.3-1844 t/s ~56.7-67.5 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.

FAQ

Can the NVIDIA DGX Spark run DistilProtBert?โ–ถ

Yes. At Q4 it generates ~1547.3-1844 tok/s and fits up to 1733k context; beyond that it spills to SSD swap and slows to ~56.7-67.5 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1733k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.