🔌
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 DDPM-IP (CelebA)?
Yes — here's how fast and up to what context, with real measured numbers.
DDPM-IP (CelebA) on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1726k | 337.8-402.6 t/s | ~12.4-14.7 (slow) |
| Q8 | yes | 1730k | 675.6-805.1 t/s | ~24.7-29.5 (slow) |
| Q4 | yes | 1732k | 1206.4-1437.7 t/s | ~44.2-52.7 (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 DDPM-IP (CelebA)?▶
Yes. At Q4 it generates ~1206.4-1437.7 tok/s and fits up to 1732k context; beyond that it spills to SSD swap and slows to ~44.2-52.7 tok/s.
How much context fits?▶
At Q4, up to about 1732k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.