// RUN — FITS

Can the NVIDIA DGX Spark run BERT-RBP?

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

BERT-RBP on the NVIDIA DGX Spark

QuantFits?Max contextSpeedBeyond context
FP16 yes 1732k 905.9-1079.6 t/s ~33.2-39.5 (slow)
Q8 yes 1733k 1811.7-2159.2 t/s ~66.4-79.1 (slow)
Q4 yes 1734k 3235.2-3855.7 t/s ~118.5-141.2 (slow)

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

→ All models on the NVIDIA DGX Spark → BERT-RBP specs → Other hardware

FAQ

Can the NVIDIA DGX Spark run BERT-RBP?▶

Yes. At Q4 it generates ~3235.2-3855.7 tok/s and fits up to 1734k context; beyond that it spills to SSD swap and slows to ~118.5-141.2 tok/s.

How much context fits?▶

At Q4, up to about 1734k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.