You are not logged in.
I am now a debian 12 user(debian 12 + windows11 pro),and recently I tried archlinux on my old laptop(single archlinux), which works well. And I what to shift my PC OS to archlinux.(archlinux + windows11 pro) .
I read a lot from archwiki about Nvidia. However, there are something I worry about, which caused my hesitation.
My Basic Device:
My CPU: i5-13400F(without integrated GPU)
MY GPU: Nvidia RTX 3060TI(G6X)
1. Is the resolution in multiscreen mode correctly displayed? I have 2 screens(2560x1440 and 1920x1080), which results a resolution of 3708x2520 in my debian 12 now. I doubt whether problems like this still exists in plasma 6 ? That really causes trouble to me in softwares like screenshot.
2.Does cuda in python and cuda from driver work well ? I sometimes use torch for my work. That means I use "conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia"(get from website of torch), however the cuda version from torch < cuda version from driver, which caused cuda unavailable on my Debian 12 before. So I am not sure. It is difficult to test it on my laptop cause it's not equipped with nvidia GPU.
Last edited by scarlet (2024-06-07 17:41:56)
Offline
1) that's mostly a xorg peculiarity that will behave the same, unless you opt to switch to wayland. How your screenshot software logically handles this in a normal xrandr session depends on your screenshot tool, I can't remember ever having specific issues with this even on xorg. Make sure you don't unnecessarily generate a xorg.config, those will often setup suboptimal defaults.
2) You probably want/should be using the repository packages here unless you are dead tied to a specific version. The packages in the repositories are correspondingly built and tested to work properly. e.g. all of the ones in https://archlinux.org/packages/?q=tensorflow-opt-cuda
Offline
Thank you so much!
Offline