Running a local LLM with two GPUs

Let's perform surgery on two old systems to combine them into one AI powerhouse, capable of running our local LLM quite well! We'll show how to set up llama.cpp to support multiple GPUs in the same system, and show a couple of ways to take advantage of our new combined VRAM pool. 00:00 Intro 02:47 Surgery results 03:30 Building llama.cpp 05:13 First test 07:02 Two models at once 09:13 Summary 09:42 Outro