🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the NVIDIA DGX Spark run ProBERTa?
Yes — here's how fast and up to what context, with real measured numbers.
ProBERTa on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1734k | 2264.7-2699 t/s | ~83-98.9 (slow) |
| Q8 | yes | 1734k | 4529.3-5398 t/s | ~165.9-197.7 (slow) |
| Q4 | yes | 1735k | 8088.1-9639.2 t/s | ~296.3-353.1 (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 ProBERTa?▶
Yes. At Q4 it generates ~8088.1-9639.2 tok/s and fits up to 1735k context; beyond that it spills to SSD swap and slows to ~296.3-353.1 tok/s.
How much context fits?▶
At Q4, up to about 1735k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.