Linux 7.3 improves performance when running out of vRAM
- ID
- 15183
- Status
- summarized
- Published
- 18 Aug 2026, 3:51 PM
- Fetched
- 20 Aug 2026, 7:09 AM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://pixelcluster.dev/VRAM-Overcommit/
- Source URL
- https://hnrss.org/best
Summary
- Score
- 4.0
- Created
- 20 Aug 2026, 8:13 AM
- Tags
- Audience
- developersai_ml_learners
What happened
Linux 7.3 will include upstreamed kernel patches by the author that improve GPU VRAM overcommit performance, primarily for gaming. The post explains the fundamental bandwidth bottleneck: PCIe 4.0 x16 caps at ~32GiB/s, meaning if more than ~1GiB of evicted memory must be fetched per frame, 30 FPS becomes physically impossible regardless of driver improvements.
Why it matters
If you run local AI models on consumer Linux GPUs and push past physical VRAM, the same PCIe bandwidth ceiling applies — ~1GiB of system-memory fetches per frame is the hard limit before performance collapses. This patch may marginally help eviction decisions, but the article is gaming-focused and doesn't address LLM/inference workloads, so the practical takeaway for AI builders is limited to understanding the bandwidth math.
Discussion angle
The PCIe bandwidth math (~32GiB/s, ~1GiB per 33ms frame) is a useful concrete ceiling to discuss for anyone trying to run models on GPUs that exceed VRAM — but the patches themselves are gaming-oriented and may not translate to inference workloads.