I'm running Llama3.2:3b in Ollama on a Pi 5 8GB. It's running well, but responses to inputs take anywhere from 6 to 50 seconds depending on the length of the input and the content memory fed to the system. I've never really thought about overclocking for anything, but this one might be worth it... is there really any significant performance increase on something like a LLM?
Statistics: Posted by dave xanatos — Tue Feb 25, 2025 2:27 am — Replies 1 — Views 64