Used RTX 3090 purchase review: Local AI performance crushes 3060, Qwen 35B 20x faster
Yanzihko · reddit · 2026-08-21
An author shared their experience purchasing a used INNO3D RTX 3090 for local AI inference.
Hardware Specs:
- Price: $1000 (considered high but the best offer found).
- Condition: Ex-mining card (paste and pads replaced), currently undervolted to 1700MHz.
Performance Comparison:
- Running Qwen 35B, the 3090 generates responses in 20 seconds, compared to over 400 seconds on the previous RTX 3060—a massive performance leap.
- The author expressed anticipation for the performance of the 4090 and future 5090.
Misc: The author asked for further optimization tips for Stable Diffusion and Ollama (beyond MSI Afterburner).
More from Infra
- Wake: macOS app unifies chat history for 13 code agents — aigclink · 2026-08-21
- Kubernetes CPU Limits Make Apps Slow and Costly: Proof and Experiments — JeremyCMorgan · 2026-08-21
- Productionizing AI Apps: OpenTelemetry, On-Call Agents, and Full Observability Workflow — Al_Grigor · 2026-08-21
- LLMRouter 2.0: Unified Infrastructure for LLM Routing Dev and Eval — youjiaxuan · 2026-08-21
- Run MiniMax H3 locally on 12GB GPUs: 15-second multi-shot ComfyUI template — vortis23 · 2026-08-21
- llama.cpp adds tensor split for LFM2/MoE, boosting inference performance significantly — pmttyji · 2026-08-21