So NVIDIA just doesn’t cut it on Linux/proton I’ve come to learn. Looking at the best bang//buck, it this the AMD card people are flocking to? 7800 XT maybe?

  • Amongussussyballs100@sh.itjust.works
    link
    fedilink
    arrow-up
    26
    ·
    4 months ago

    I cannot speak for this card itself, but moving from Nvidia to AMD made my life so much easier. Wayland works a treat, and updates never leave me with a black screen from silly diver issues. However anything for local llms is a massive pain in the ass to use compared to Nvdias cuda, rocm is quite half-baked.

    • abcdqfr@lemmy.worldOP
      link
      fedilink
      arrow-up
      4
      ·
      4 months ago

      I’ll definitely be keeping my nvidia card for ai/ml /cuda purposes. It’ll live in a dual boot box for windows gaming when necessary (bigscreen beyond, for now). II am curious to see what 16gb of amd vram will let me get up to anyway.

  • circuitfarmer@lemmy.sdf.org
    link
    fedilink
    arrow-up
    12
    ·
    4 months ago

    Switched everything over to AMD and have never needed to look back. It is way more It Just Works on AMD.

    Love, the Steam Deck

  • Sanctus@lemmy.world
    link
    fedilink
    English
    arrow-up
    9
    ·
    4 months ago

    I have had a shit time with my 2080 TI. If I had the money I’d jump for an 7800 XT in a heartbeat.

  • rem26_art@fedia.io
    link
    fedilink
    arrow-up
    6
    ·
    4 months ago

    I’ve got a 7800XT now and I moved from a 1070 and I’ve been happy with it overall. I’m on Fedora and I bought the 7800 kinda close to launch, so I went through some issues that seem to have been solved by now. Nothing that really made me go “gee I wish I hadn’t switched”.

    I don’t do anything related to streaming, or machine learning, so I can’t really speak to it’s ability with those, but gaming has been stable, and, aside from a now solved problem with rocm, it works fine with Blender cycles (at least on Fedora 40). Davinci Resolve has worked fine too. On launch, there wasn’t VAAPI support for AV1, but that works just fine for me now. (VAAPI is the open source interface for GPU video acceleration).

    Currently, I’d say the experience is perfectly fine.

  • cybersandwich@lemmy.world
    link
    fedilink
    arrow-up
    6
    ·
    4 months ago

    I switched from Nvidia for amd for the same reason: “and is better on Linux”.

    In my experience you are just making different tradeoffs. I use pop so your mileage may vary but Nvidia was easy to use and upgrade. It’s not nearly as bad as people let on.

    AMD on the other hand isn’t as seamless as people let on. And the open source drivers, while awesome, don’t let you take advantage of the codecs for video streaming or even alot of the AI ML stuff, so you switch to the proprietary drivers and they are slightly buggy.

    I wish I kept my 3070ti over the 6900xt.

    Unless they figure out a way to let me use av1 or rocm more easily then my next card will be Nvidia again.

    • AProfessional@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      ·
      edit-2
      4 months ago

      Video decoding/encoding should work fine, better than Nvidia as fewer things support nvdec (the vaapi wrapper is enough though).

  • AmbiguousProps@lemmy.today
    link
    fedilink
    English
    arrow-up
    6
    ·
    4 months ago

    Yes. The nvidia drivers on linux are horrible, and always have been. Since I ditched my nvidia 2080 it’s been much more stable.

  • HumanPerson@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    6
    ·
    4 months ago

    IMO, intel has underrated linux drivers. You get solid 3d, codecs, compute, etc. ootb. Assuming your distro supports it. You may be looking into something higher end, though.

      • HumanPerson@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        4
        ·
        4 months ago

        I daily drive arc on linux. They’re not as bad as people say. Not fully there, but opencl support requires one package that is in most distros repos, same for video. Not saying they’re perfect, or even better than amd, but they are a lot better than people seem to think.

  • thayer@lemmy.ca
    link
    fedilink
    English
    arrow-up
    4
    ·
    edit-2
    4 months ago

    We swapped our 2080s for 6800 XTs last year and couldn’t be happier…a 7900 GRE should be great for gaming, but I can’t speak to LLM performance.

  • faede@mander.xyz
    link
    fedilink
    arrow-up
    4
    ·
    4 months ago

    I have Nvidia and now that explicit sync is out my system runs incredibly smoothly. Every game I have tried works great.

  • boredsquirrel@slrpnk.net
    link
    fedilink
    arrow-up
    4
    arrow-down
    2
    ·
    4 months ago

    In my experience, as a nongamer just laptop user, Intel is way more stable than AMD too. Might consider an Intel GPU? But I only know the integrated ones on Laptops, which work really well

    • cmnybo@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      6
      ·
      4 months ago

      The dedicated Intel GPUs have nowhere near as much performance as an RX 7900. The video encoder is very good though.

      • boredsquirrel@slrpnk.net
        link
        fedilink
        arrow-up
        2
        ·
        4 months ago

        Also stuff like waking up from suspend, random freezes and power management work better than on AMD. I would assume this also applies to the Arc GPUs?

        • Vik@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          4 months ago

          Can’t say I’ve experienced any such issues on recent AMD UMA or dGPU systems; power management is pretty well documented and generally reliable without any need for user intervention. Curious as to which platforms were problematic in your experience for my own learning; it’s likely that anything pre-polaris was kind of wonky

          I’ve also not had any issues with Intel integrated graphics on Linux, but ANV on Arc is a bit messy with translation layers like vkd3d right now. The gen12 (DG1, RKL and later) driver & technology stack appears to be quite different.

          • boredsquirrel@slrpnk.net
            link
            fedilink
            arrow-up
            1
            ·
            4 months ago

            No idea but a Thinkpad T495 with AMD Vega mobile graphics. Dont remember the generation.

            Fedora + systemd + KDE Plasma had always issues waking up from sleep. Needed to force shutdown a lot.

            • Vik@lemmy.world
              link
              fedilink
              English
              arrow-up
              1
              ·
              edit-2
              4 months ago

              huh, interesting. Picasso (3000)? Raven Ridge (2000)?

              Similar if not practically identical chassis to my T14 Gen2a. My one is Cezanne based and the experience with fedora/wayland/gnome has been perfectly smooth. Did you spot any logs relating to those issues?

              • boredsquirrel@slrpnk.net
                link
                fedilink
                arrow-up
                1
                ·
                4 months ago

                Picasso.

                No logs afaik. But that pile of crap doesnt even boot anymore, like no BIOS. Trying to fix it somehow.

  • d-RLY?@lemmy.ml
    link
    fedilink
    arrow-up
    2
    ·
    4 months ago

    I am all AMD both PC (currently Windows but have used Linux on systems with AMD and Nvidia over the years) and Steam Deck (of course). AMD is overall easier. That being said, Nvidia is supposedly in process of making opensource drivers. I believe they are going to be focusing on their newer cards. So it might be worth researching into any recent news on their progress. Always good to have options if you get a better deal on one vs another.

    • Sammirr@aussie.zone
      link
      fedilink
      arrow-up
      7
      ·
      4 months ago

      Worth noting that Nvidia only intends to open source the kernel driver. This is only half the driver, as a userspace blob will still be required, and that will remain closed and proprietary.

      • d-RLY?@lemmy.ml
        link
        fedilink
        arrow-up
        1
        ·
        4 months ago

        Ah, good to know. Though wouldn’t it still at least make it easier for people to install overall? The last time I messed with Nvidia on Linux gave me issues even with using a named supported distro on their site. Would error out about “missing headers” or something like that. Given this was many years ago and before distros would offer an Nvidia specific iso. Mostly just curious in the event that I needed to help someone that is all-in on having one of their cards.

        • Sammirr@aussie.zone
          link
          fedilink
          arrow-up
          2
          ·
          4 months ago

          Hard to know. Will the interface be specific to driver versions? Will it require an updated kernel driver for each userspace driver as it does now? I don’t know that we have the answers.

  • Vik@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    edit-2
    4 months ago

    Primarily use AMD graphics. My key issue on linux is the GPU reset situation, which can make experiences like VFIO and LookingGlass less than optimal, though there seems to be some commitment from all IHVs to improve desktop expeirence under Linux.

    AMD and Valve work fairly closely on such endeavours which is neat, though we also have nvidia getting their shit together for Wayland and now offering an open kernel module (even in lieu of open, first party UMDs, for which the NVK driver is well worth investigating).

    The open kernel module and NVK driver are applicable to Turing (your current GPU) and newer. Check them out.

    You don’t owe it to anybody to align to a given vendor, just use what works best for you at a price point that doesn’t suck. If there are any specific use cases you’d like to know about then I’m sure the people here would be happy to test and report back.