• 0 Posts
  • 42 Comments
Joined 2 years ago
cake
Cake day: October 10th, 2023

help-circle
  • Thanks. Seems you’re correct and Sunshine profits from CUDA when using this NVFBC but it doesn’t seem to explicitly depend on it to run. I tried it pretty soon after installing EOS and can’t imagine having installed CUDA so early. Last time I explicitly did it was for Ollama.

    I just looked on that EOS machine on which Sunshine worked right from the start and no CUDA package is installed but nevertheless nvidia-smi shows CUDA Version 13. This seems to be because the nvidia driver brings the CUDA runtime with it which differs from the big CUDA toolkit. I’ll try CachyOS again when I eventually buy an AMD GPU.

    I think you meant Cachy when you pointed out Limine is an option while doing automatic partitioning. And yes this worked very good. I meant that Endeavour doesn’t offer it at that point in install.

    Have a nice week.




  • Recently tried CachyOS on my HTPC. Couldn’t get Sunshine to run. It looked beautiful though and the docs seemed really in-depth. Endeavour (thanks to Arch, I know) is just so stable and fast. Just the right amount of pre-installed stuff like reflector, yay, the firewall-config app for firewalld with sane defaults, nice BTRFS subvolume layout, correct NVidia drivers. Would be nice to have Limine (for BTRFS snapshots) as an option besides SystemD boot and Grub in Calamares but I installed it in addition to SystemD boot. Gaming works tremendously good. Everything else too.

    I really can’t recall the last time an update went wrong. On both of my machines (one Intel, one AMD, both NVidia).



  • A recent famous case that comes to mind is when some independent researchers found an ex-RAF terrorist on the run (or call her whatever your political inclination suits) via facial recognition of photos on social media of a capobeira club. I assume the reference material were some age old photos as she was on the run for decades.

    I don’t know much about facial recognition but I only heard it works damn well and considering OS and phone manufacturers employing it and are confident enough to not have their users locked out or compromise security let’s me assume that it works damn well.











  • Don’t know about your hardware. I don’t own a notebook anymore. I read good things about the AUR package optimus-manager-qt for hybrid GPUs (iGPU+dedicated GPUs) but also that it can be a bit tricky.

    I exlusively used dedicated Nvidia cards in desktop rigs with Arch & EndeavourOS since 2017 when I switched from Win 10. Additionally exclusively KDE.

    Though I had a bit of experience with other distros and desktop environments before my switch I’d wager to say you should give one last try to EndeavourOS, even if you have barely any Linux experience. I mean you had so many failed attempts. One more won’t hurt.

    Use EndeavourOS not arch. First, it uses the standard initial graphical system-setup (Calamares), then it comes with some good default settings & tools and finally a welcome screen which features links to additional tools like mirror selection (for faster updates), update shortcuts, package search, docs/wikis/forums or logs.

    I’d select KDE in Calamares and I’d install the graphical package manager octopi via “yay octopi” after system installation and activate yay for the AUR in the octopi settings as e.g. optimus-manager-qt (which you should only use with hybrid GPUs) is only available in the AUR. You need to click the alien symbol in octopi to install from the AUR.

    The AUR (Arch User Repository) is the repository for packages not available in the main repositories. AUR packages are user contributed where the maintainers write a so called PKGBUILD file which contains the steps to build and install a package from foreign sources (e.g. from a debian DPKG or from github sources). With octopi you can quickly open the PKGBUILD file and look from where the maintainer pulls the parts of the package.

    The amount of software available in the AUR is gigantic but it can potentially contain malware (which happened a very few times). But you’ll have a hard time finding users who actually had that happen to them. A good indicator that the package is ok are its number of votes. But if you really want to know you have to check the sources in the PKGBUILD. If they come from github, you could check the github-repo and only it’s stars (votes) if you won’t read the sourcecode.


    That all sounds mighty complicated but it isn’t. Just try to install packages from the main repo. Click the alien symbol only when you don’t find something official.

    So with octopi and the welcome screen you don’t need to enter any terminal commands for package installation or the system update. I had only a few updates where problems occurred in like 7 years and they were always fixable. The Arch Wiki and the Endeavour forums could always help.

    I can’t guarantee you’ll have a better experience than with the other distros and you will meet some bumps or roadblocks for sure. I’m not playing the the most current games and a lot of retro games via Lutris and Heroic. For some of them I had to tinker a bit and try different starters than Steam. Arma, Path of Exile, Sekiro (fitgirl repack), Diablo Immortal were tricky but all the steam games or e.g. Witcher 3 via Heroic run very nice.

    On the screen where you login (usually SDDM) you can switch between Wayland and X11. Which are two very different Display managers. Wayland is the replacement for the very old X11. It works way(land) better with AMD GPUs than with Nvidia which are usable though but work much better on X11. Games can be faster on wayland for Nvidia than on X11. But things like missing color management in nvidia-settings make me stay with X11.




  • I watched one video and read 2 pages of text. So take this with a mountain of salt. From that I gathered that deepseek R1 is the model you interact with when you use the app. The complexity of a model is expressed as the number of parameters (though I don’t know yet what those are) which dictate its hardware requirements. R1 contains 670 bn Parameter and requires very very beefy server hardware. A video said it would be 10th of GPUs. And it seems you want much of VRAM on you GPU(s) because that’s what AI crave. I’ve also read 1BN parameters require about 2GB of VRAM.

    Got a 6 core intel, 1060 6 GB VRAM,16 GB RAM and Endeavour OS as a home server.

    I just installed Ollama in about 1/2 an hour, using docker on above machine with no previous experience on neural nets or LLMs apart from chatting with ChatGPT. The installation contains the Open WebUI which seems better than the default you got at ChatGPT. I downloaded the qwen2.5:3bn model (see https://ollama.com/search) which contains 3 bn parameters. I was blown away by the result. It speaks multiple languages (including displaying e.g. hiragana), knows how much fingers a human has, can calculate, can write valid rust-code and explain it and it is much faster than what i get from free ChatGPT.

    The WebUI offers a nice feedback form for every answer where you can give hints to the AI via text, 10 score rating thumbs up/down. I don’t know how it incooperates that feedback, though. The WebUI seems to support speech-to-text and vice versa. I’m eager to see if this docker setup even offers APIs.

    I’ll probably won’t use the proprietary stuff anytime soon.