Is Running Local LLMs on SSD RAID0 Viable?
KeinNiemand · reddit · 2026-07-17
A Reddit user asks if it's possible to use multiple SSDs in a RAID0 setup to run LLMs, acting as a substitute for expensive VRAM and system RAM.
The idea is to leverage the much lower cost-per-gigabyte of SSDs compared to RAM. By using 16–32 high-speed SSDs in RAID0, the theoretical bandwidth could approach memory-level speeds. However, the user acknowledges potential issues like higher latency, the overhead of constantly moving data between SSD/RAM/VRAM, and insufficient PCIe lanes on the motherboard.
Essentially, the post discusses whether bandwidth can compensate for latency disadvantages in local LLM inference, and if this "SSD stacking" approach is a truly affordable alternative for running local models.
More from Infra
- NeurIPS 2026 workshop calls papers on on-device intelligence — YiMaTweets · 2026-07-21
- Milled from Solid Aluminum: AI Rig Multi-GPU Case for Local Compute — dee_hw · 2026-07-21
- FutureCaribbean’s Buildathon offers $50K, H200 compute, and an NYSE pitch — HeyAmit_ · 2026-07-21
- A new series tests which data-science workflows can run on GPUs today — pandeyparul · 2026-07-21
- Former AWS operator says Bedrock margins can beat SageMaker as agentic AI lifts CPU demand — RihardJarc · 2026-07-21
- Engram shows how agent memory can keep, rewrite, or delete facts asynchronously — philipvollet · 2026-07-21