Writing · Tag

#multi-gpu

1 post tagged #multi-gpu.

Jun 8, 20267 min read
llama.cpp --tensor-split: Run a 70B Across 2 GPUs (2026)
Your 70B model won't fit on one GPU. Here's how llama.cpp --tensor-split, --main-gpu, and --split-mode spread it across two cards on a real consumer rig.
#llama.cpp#local-llm#gpu#multi-gpu

The AI agent build notes

Real costs, real tools, no fluff. M-F when I ship, publish, or learn something worth sending.