← sources

Guide on setting up new pc from scratch for local ai model

source: reddit (reddit:LocalLLaMA)published 2026-11-13 16:10ingested 2026-11-14 02:42

retained text

I’ll include my specs at the end. Because probably no one’s going to read it..

For my new pc im considering migrating to Linux ubuntu, and set up my llama.cpp from there, my old pc had llama.cpp Vulkan sdk running on windows 11

Im looking for any guidance or tips while setting them up to ensure I didn’t miss any critical steps that could end up loosing a bunch of tps or something worse

Beginner mistakes i would say

I do know one tips and it’s plugging your hdmi cable to the integrated graphics card, that way it saves vram. But I don’t have one..

I appreciate any advise

Bought a B550-A-Wifi ii  
Rtx3060 1x (plan on getting another one in 2-3 months)

32gb ddr4 3200 2x16 (plan on getting 32gb more, total 64, in 4months) **saving up**

And a 750W cooler master v3 gold modular PSU (not sure if it’s suitable for my small server  
R7 5700x (no integrated graphics card)

document fetched 2026-11-14 00:46 · canonical https://reddit.com/r/LocalLLaMA/comments/1xvbwl9/guide_on_setting_up_new_pc_from_scratch_for_local