• Pennomi@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    ·
    21 days ago

    It depends. A lot of LLMs are memory-constrained. If you’re constantly thrashing the GPU memory it can be both slower and less efficient.