I’m on a GNU/Linux and I have a second gfx card that is nvidia but not plugged in to anything other than my Motherboard, can I use this still for cuda based ai stuff?

  • what@lemmy.world
    link
    fedilink
    arrow-up
    11
    ·
    1 year ago

    I’ve been using cuda for cracking hashes without issue. Now if someone at Nvidia could work on not making their driver suck at daily driving on Linux it would be great

  • PeterPoopshit@lemmy.world
    link
    fedilink
    arrow-up
    3
    ·
    edit-2
    1 year ago

    Actually I got an nvidia card working on easy diffusion on Debian. The barrier for getting a text chat ai working with gpu acceleration is actually the fact that I don’t have the patience to deal with all that python venv nonsense so I use llamacpp. It runs in c++ which means no python dependencies to fuck you with at the cost of slower cpu-only generation.

    Easy Diffusion just happens to be simple enough that I could actually figure out how to get it to work (it’s in python and needs a virtual environment) but it’s a different story for the text ais.

    If you actually had the patience and knowledge to deal with all the python issues and/or had a distro that makes it easy (different distros deal with pip differently), I don’t doubt you’d be able to get Nvidia card acceleration working on some text chat ai.

  • Dale@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    I have been on my laptop with 4Gb nvidia gpu. If you’re using the webgui there’s optimization parameters like xformers and I think opt.sdp.attention or something that uses about half the amount of memory. I had to update the graphics card to get it working first.

  • Enzy@lemm.ee
    link
    fedilink
    Svenska
    arrow-up
    2
    arrow-down
    1
    ·
    1 year ago

    Yes just rotate it and connect it to the cpu bearings