Repurpose an old low-VRAM GPU just for mmproj in llama.cpp — an order of magnitude faster

inthesearchof · reddit · 2026-09-12

A practical llama.cpp tip: the common --no-mmproj-offload trick saves VRAM but is painfully slow, especially for agentic coding with vision. Instead, dedicate a spare low-VRAM GPU solely to the vision projector via --mmdev CUDA1 (your GPU index). The author reports this is roughly an order of magnitude faster than --no-mmproj-offload with no impact on main-model inference speed.

Original post →

More from coding & agent

coding & agent channel →