Exploring a Dual 3080Ti Local Swarm Setup to Accelerate Qwen Inference

Forward_Jackfruit813 · reddit · 2026-08-29

A user proposes building a local inference cluster using a second, cheap 3080Ti GPU to speed up their workflow. The setup involves a gaming desktop (Ryzen 5700X3D) and a dedicated AI PC (96GB RAM) currently running Qwen4 Flash Next slowly. The plan is to run Qwen 3.8 27B workers on the dual-GPU desktop, using Kimi Code CLI's swarm feature to serve the AI PC over LAN. The user seeks advice on whether the performance gains justify the cost of a new motherboard and PSU.

Original post →

More from Infra

Infra channel →