Redditor builds 8x R9700 local inference rig with Threadripper and dual 1600W PSUs
NunzeCs · reddit · 2026-10-28
A Reddit user showcased an 8x R9700 local inference rig: Threadripper 9955WX, 128GB DDR5-5600 (256GB deemed too expensive), Asrock WRX90 WS EVO, and dual Seasonic PSUs (2200W + 1600W), with a cheap AliExpress GPU mount and three fans on top for exhaust.
The upgrade from 4 to 8 GPUs leaves headroom for bigger models. Currently running QFN quantization — after trying GLM 5.3 Flash, they find QFN with radiance the better experience. The bifurcation card plus 6 riser cables still has issues; a retimer is on the shopping list.
More from Infra
- Celestica guides 2027 revenue to $40B, up 90%, with margins still expanding — TiernanRayTech · 2026-10-28
- RTX 5090 runs a 33B Q4 model at ~86 tokens/s: buy memory bandwidth, not TFLOPS — AIFlow_ML · 2026-10-28
- Model response speed is mostly memory bandwidth, not compute — and Rubin regressed — AIFlow_ML · 2026-10-28
- NVIDIA Vera Rubin NVL72 page goes live: 72 Rubin GPUs, rack-scale agentic AI supercomputer — AIFlow_ML · 2026-10-28
- SK Hynix scraps US IPO plans for NAND subsidiary Solidigm, WSJ reports — Beth_Kindig · 2026-10-28
- NVIDIA quietly cut Vera Rubin's HBM4 bandwidth 13% per GPU — the spec that matters for inference — AIFlow_ML · 2026-10-28