DGX Spark driver fixes reclaim 64GB of 'missing' GPU memory and boost page faults 49x
QuixiAI · x · 2026-09-06
QuixiAI (Eric Hartford) shares two fixes for the NVIDIA open kernel driver, both verified on a DGX Spark (GB10):
- Memory reclaim: freed GPU memory now returns to the OS when a process exits — previously 64 GB went 'missing' and vLLM refused to start.
- Huge pages: GPU page faults on system memory now use huge pages, raising first-touch bandwidth from 0.4 to 19.6 GiB/s (49x).
More from Infra
- How much VRAM do you actually need for local document AI like paperless-ai? — pjdonovan · 2026-09-06
- 96-plasmid yeast transformation in 3.25 hours with open-source Pylabrobot — nlarusstone · 2026-09-06
- Dev Claims Fix for NVIDIA's Intentional P2P Nerf on Consumer GPUs — QuixiAI · 2026-09-06
- Fine-tuning LLMs in the browser: WebGPU training PoC built on llama.cpp — ngxson · 2026-09-06
- Ollama cloud launches off-peak token pricing: DeepSeek-V4 at half price — ollama · 2026-09-06
- Ollama cloud full price list: $0.015 to $15 per million tokens across models — paw_lean · 2026-09-06