🔌
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 ImageBind?
Yes — here's how fast and up to what context, with real measured numbers.
ImageBind on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1706k | 106.9-127.4 t/s | ~3.9-4.7 (slow) |
| Q8 | yes | 1721k | 213.8-254.8 t/s | ~7.8-9.3 (slow) |
| Q4 | yes | 1727k | 381.8-455.1 t/s | ~14-16.7 (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 ImageBind?▶
Yes. At Q4 it generates ~381.8-455.1 tok/s and fits up to 1727k context; beyond that it spills to SSD swap and slows to ~14-16.7 tok/s.
How much context fits?▶
At Q4, up to about 1727k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.