Forget megabucks Nvidia GPUs, apparently all you need to run an LLM is a Pentium II CPU from 1997

⁨15⁩ ⁨likes⁩

Submitted ⁨⁨1⁩ ⁨month⁩ ago⁩ by ⁨cm0002@lemmy.world⁩ to ⁨technology@lemmy.zip⁩

https://www.pcgamer.com/software/ai/forget-megabucks-nvidia-gpus-apparently-all-you-need-to-run-an-llm-is-a-pentium-ii-cpu-from-1997/

Comments

Sort:hotnew top

AmbiguousProps@lemmy.today ⁨1⁩ ⁨month⁩ ago
The other catch is that the model is tiny, which will cause hallucinations to skyrocket.

source
- projectmoon@forum.agnos.is ⁨1⁩ ⁨month⁩ ago
  @ambiguousprops@lemmy.today still an impressive feat, though. Surprised that anything resembling modern software even runs on a system that old now. Probably requires an ancient kernel version and other things. Is there anywhere where they detail this process more specifically? (I only had time to skim the article)
  
  source
  - AmbiguousProps@lemmy.today ⁨1⁩ ⁨month⁩ ago
    The GitHub has more details. github.com/exo-explore/llama98.c
    
    source