Can a $351 mini PC really run a 30B+ AI model locally? π
We tested the MINISFORUM UM790 Pro with Qwen 35B and Gemma models β and the results are surprisingly solid.
π» Setup:
β’ AMD Ryzenβ’ 9 7940HS
β’ Radeon 780M iGPU
β’ 48GB DDR5 RAM (dual-channel)
β’ 512GB SSD
βοΈ What we did:
β’ Increased UMA frame buffer
β’ Installed Adrenalin Edition & dependencies
β’ Ran llama.cpp with local LLMs
π Result:
Smooth local inference with 20+ tokens/sec (real test, may vary by setup)
No cloud. No API fees. Just local AI. π
π Learn more: https://s.minisforum.com/MiniPC
#MINISFORUM #UM790Pro #LocalAI #MiniPC #LLM #AI #Tech
source
