Dual RTX 3060 vs single RTX 3090: choosing a $500 vs $1000 build for a home local LLM server
Bakkario · reddit · 2026-09-08
A user plans to sell their 3060 6GB laptop and build a 24/7 home private inference server, torn between two options:
- Dual RTX 3060 12GB ($500): split models across cards over PCIe Gen4, paired with 32GB DDR4 RAM — the budget route
- Single RTX 3090 24GB ($900-1000): more VRAM and bandwidth, but roughly double the cost
They're asking experienced builders how much performance they'd leave on the table with the dual-3060 route, a classic cost-performance dilemma in local LLM deployment.
More from Infra
- Data center buildout is driving freight demand that the Cass index misses — kernelangus420 · 2026-09-08
- Strix Halo users ditch official llama.cpp: optimized forks hit ~60 t/s decode vs ~20 t/s — feelspeaceman · 2026-09-08
- AMD's Halo Station Touts 2.6TB Memory, Tripling NVIDIA's 748GB DGX Station, Shipping 2027 — shashib · 2026-09-08
- DavidSHolz: AI services should let users pay extra to spin up CPU/GPU VMs — DavidSHolz · 2026-09-08
- What broke running an agent fleet 24/7 — it wasn't prompt quality — ParrotIntegrated · 2026-09-08
- Huawei 950DT reportedly priced at ~$16K per chip in DeepSeek order, below expectations — teortaxesTex · 2026-09-08