Mixing RTX 3060 and 5060 Ti for local 30B model inference?

MkGod · reddit · 2026-08-18

A user discusses the feasibility of combining an RTX 5060 Ti (16GB) with a used RTX 3060 (12GB) to achieve 28GB VRAM for running 30B+ models locally. Key discussion points include:

Original post →

More from Embodied

Embodied channel →