Multi-GPU PCIe Switch Setup Shows Almost No Performance Loss
Old_Grapefruit8774 · reddit · 2026-07-11
The author tested a direct-connect setup for 6 MI50 GPUs on an X99-E-WS motherboard, alongside a hybrid setup using a PEX8749 PCIe switch (4 cards routed through the switch chip, 2 directly connected).
Test Configuration
- Asus X99-E-WS, flashed with a custom BIOS supporting multi-GPU
- Xeon E5-2680 v4
- 128GB DDR4
- Goal: Provide around 1TB VRAM for multi-machine/multi-GPU local AI inference
Results
After multiple runs using llama bench:
- The tg metric typically improved by 0.8%–2.8% under the switch setup
- The pp metric saw less than a 1% loss
- Overall differences across multiple test combinations were minimal, with some scenarios breaking even
Conclusion: On this older platform, hanging部分 GPUs under the PEX8749 switch didn't cause a performance disaster and even yielded slight gains in tg.
More from coding & agent
- Data engineering, not agent frameworks, is the real bottleneck for enterprise AI agents — dhruv2038 · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- 105 hidden bugs, 2 repos: DeepSeek V4.1 Flash fixes 24 at $1.80 vs Opus 5's 27 at $51.33 — ChartsJournalX · 2026-09-11
- Investment Analyst Asks How to Build a Claude-Based Diligence Agent Stack — Careless_Tie2286 · 2026-09-11
- Treating agents like 50 First Dates: a 3-layer context system so every conversation doesn't start from zero — evielync · 2026-09-11
- Running the Firefox MCP on Android via Termux, ngrok, and mcp-proxy — Nervous-Strain7544 · 2026-09-11