Dual RTX 3060 12GB performance for Qwen3.8-27B
Mean-Ad1493 · reddit · 2026-08-23
Plans to add a second RTX 3060 12GB via a PCIe riser to run Qwen3.8-27B locally using layer-split mode. Requests real prefill/decode token speeds to set expectations, noting that upgrading to a 3090 is not an option due to budget constraints.
More from Infra
- DeepSeek V4 Flash 75% Off on Merge Gateway — shensi · 2026-08-24
- Antirez explains Speculative Decoding sampling mechanism — antirez · 2026-08-24
- Hot Chips Analysis: Why There Is a Memory Shortage — firstadopter · 2026-08-24
- Φ-Bench: Can LLMs Engineer the Infrastructure That Powers Them? — 青稞AI · 2026-08-24
- Opinion: Data centers generate $27B in tax revenue, yet localities keep banning them — robleclerc · 2026-08-23
- How to check GPU memory row-remapping health for extended lifespan — StasBekman · 2026-08-23