🔌
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 CLIP (ResNet-50)?
Yes — here's how fast and up to what context, with real measured numbers.
CLIP (ResNet-50) on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1732k | 1124.7-1340.3 t/s | ~41.2-49.1 (slow) |
| Q8 | yes | 1734k | 2249.3-2680.7 t/s | ~82.4-98.2 (slow) |
| Q4 | yes | 1734k | 4016.6-4787 t/s | ~147.1-175.3 (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 CLIP (ResNet-50)?▶
Yes. At Q4 it generates ~4016.6-4787 tok/s and fits up to 1734k context; beyond that it spills to SSD swap and slows to ~147.1-175.3 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.