Old servers are cheap, especially the memory meaning you can get a lot of RAM at a low cost. Does this make it viable to run AI language models at a low cost?

Ollama:
https://github.com/jmorganca/ollama

Issue relating to AVX:
https://github.com/jmorganca/ollama/i...

PDF for the server I have:
https://www.bargainhardware.co.uk/med...