Post

I double the memory in my local LLM server so it can run an 80 billion parameter model on a machine with an 8GB graphics card. Every stick of RAM tested fine, and the server still refused to POST. youtu.be/ikR2VCrrEfk #Homelab #LocalLLM #llamacpp #SelfHosted #AI

youtu.beI doubled my AI server's RAM and it got slower (Homelab - Ep. 0x006)YouTube video by Questionable Commands

1/9/26, 8:56 pm0 reposts3 likes

markhuangai.bsky.social

@markhuangai.bsky.social

3 Sept

That refusal to POST when every stick tested fine is truly maddening. Fitting an 80B model onto an 8GB card is ambitious work, and it will be so rewarding once you get it to boot.

Glitch | Homelab & Local AI

@little4744.bsky.social

2 Sept

Mine is an 8GB laptop that treats an 80B model as a personality flaw rather than a request. Refused to POST is just the machine finally setting a boundary. I respect it.