One GPU, many VMs: a practical guide to passthrough, vGPU, and API remoting
LoganGrasby · x · 2026-07-29
A practical breakdown of the main ways to share one GPU across many VMs:
- Passthrough — one GPU directly attached to one VM; fastest, but no sharing.
- vGPU / mediated passthrough — hardware-level slicing such as NVIDIA vGPU or MIG; near-native speed, but needs specific datacenter hardware and licensing.
- API remoting — GPU calls are intercepted in software and forwarded to a host process; slower at about 70%–85% of native, but gives the host full visibility and orchestration over calls.
- Emulation — pure software compatibility layer; broadly compatible, but unusable for real compute.
The author says they picked API remoting for small machines because it allows orchestration that hardware slicing cannot.
More from coding & agent
- Dan Grover built a Stream Deck plugin for showing AI agents in Herdr — DanGrover · 2026-07-29
- Local lip-reading demo turns a mouthed prompt into Claude code — breath_mirror · 2026-07-29
- GitHub repo bundles battle-tested settings for Claude Code, Codex CLI and Cursor — tom_doerr · 2026-07-29
- A setup screen compares Codex, Claude, Copilot, Cursor, and Grok Build — draginol · 2026-07-29
- Beginner Asks How to Build a Local AI Project Manager for Creative Projects, Community Offers Advice — koreanalleyarcade · 2026-07-29
- Indie builders should stop micromanaging code and start managing AI agents — alexmacgregor__ · 2026-07-29