Running Qwen 35B at 82 tok/s with a mismatched 5070 Ti + external 3070 build
hereforcheesee · reddit · 2026-10-18
The author mounted an old RTX 3070 outside the case on a PCIe riser (x1 is enough since layer-splitting passes little data) alongside a 5070 Ti.
- Ollama/llama.cpp splits Qwen 3.6 35B-A3B (MoE, 3B active params) across the mismatched GPUs
- Result: 82 tok/s with 16k context
- Caveats: 23% of layers spill to system RAM, split ratio untuned, 4GB of 3070 VRAM unused
- Next step: a cheap external enclosure for the 3070
More from Infra
- SGLang Summit sponsors jump from 11 to 36 in one week — ying11231 · 2026-10-18
- Oracle CDS hits record 300bps, called first AI trade domino after Barclays sell downgrade — arthurcolle · 2026-10-18
- Agent Coding Workflow Would Cost ~$40k/Month on API Pricing, Dev Reveals — ChrisRackauckas · 2026-10-18
- PS5LM: llama.cpp port turns a jailbroken PS5 into a home LLM server (500+ GitHub stars) — wateriscoding · 2026-10-18
- DFB laser heterogeneously coupled to LNOI: 170 mW output, 45 mA threshold — jwt0625 · 2026-10-18
- China building ~69 power semiconductor fabs across 58 suppliers — zephyr_z9 · 2026-10-18