Announcement

Collapse
No announcement yet.

AMD Quietly Working On New Linux GPU Driver Support Block By Block

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by r1348 View Post
    How is your vendor lock-in going?
    Pretty good. In the two decades since I first switched to Linux (Literally. I switched from an ATi Rage 128 that would reliably crash its Windows drivers when I ran aMule to nVidia shortly after switching to Linux in 2001), I've only ever needed to apt-add-repository a PPA for a non-default nVidia driver version twice... and one of those was for a memory leak that didn't get along with my preference for leaving my session logged in for months at a time.

    Heck, concern about having to update or hold back my whole kernel just for a fix to the driver reliability regressions I hear people mention in here is one of the reasons that, if I'm in a situation where I don't need GPU compute and try AMD, I'll be buying from a local brick-and-mortar so I don't have to pay shipping to return it if I don't like it.

    (Yeah, I'm one of those people who will probably stay off Wayland until crash recovery is not only implemented, but properly mature. I remember how unreliable automatic session restore on restart is under X11.)

    ...plus, I'm still back on a GeForce GTX750... the first generation to support three monitors and the last to have unsigned firmware... I just hope it'll last until prices come down so I don't need to go digging for that bookmark for high-grade angled SATA cables and swap in my younger brother's retired GTX760 or his friend's equally large retired ATi card to keep things working.
    Last edited by ssokolow; 18 February 2022, 08:32 PM.

    Comment


    • #22
      Let's hope they're not transitioning from colorful fish to invertebrates.

      Also, very very weasel of AMD to release ROCm 5.0 with RDNA2 support, but the only RDNA2 products supported are the overpriced workstation cards.

      It is weird. Such good management of the CPU side of the business but such duplicity on the GPU side.

      Comment


      • #23
        Originally posted by horizonbrave View Post
        anyway this won't stop idiots Linux users buying NVIDIA based laptops (I guess the laptop space is the only place where you can buy a discrete graphic card at a normal price) when there are perfectly fine AMD based gaming/content_creation machine out there (finally! not that many, but they're that good).
        By the way, Hi Nvidia user, please don't feel offended...

        PS: a totally OT rant, sorry
        I had searched high and low for an AMD laptop with a discrete GPU that wasn't an nvidia... companies claim they exist, but damned if I could find any in stock. Even System76 Sells AMD CPU/Nvidia GPU laptops.

        Comment


        • #24
          Originally posted by ssokolow View Post

          How's that CUDA support going? </devils_advocate>
          Easy: just make sure you don't need CUDA or find an alternative that doesn't require that shit

          Comment


          • #25
            Originally posted by leech View Post

            I had searched high and low for an AMD laptop with a discrete GPU that wasn't an nvidia... companies claim they exist, but damned if I could find any in stock. Even System76 Sells AMD CPU/Nvidia GPU laptops.

            Maybe search left and right next time
            Dell and MSI also have something, and you can even settle for second hand ones eventually

            Comment


            • #26
              Originally posted by ssokolow View Post

              Pretty good. In the two decades since I first switched to Linux (Literally. I switched from an ATi Rage 128 that would reliably crash its Windows drivers when I ran aMule to nVidia shortly after switching to Linux in 2001), I've only ever needed to apt-add-repository a PPA for a non-default nVidia driver version twice...
              The issue with their method isnt just whether you need to add a PPA etc. It is also that until now (or maybe still), you were forced to use legacy software versions and if you tried newer software that broke, you wouldnt blame the dodgy driver, but free software.

              Nvidia do provide features that people love, but those same features are often designed in a way to lock you in. If they could get away with Apple level policies, they would.

              Comment


              • #27
                Originally posted by hoohoo View Post
                LAlso, very very weasel of AMD to release ROCm 5.0 with RDNA2 support, but the only RDNA2 products supported are the overpriced workstation cards.
                Luckily official support and things that work are 2 pairs of shoes. 5.0 is the first release where my 3500 APU, my Ryzen 4700 APU as well as my 5600 and my Radeon 5500 GPU working with OpenCL (and HIP for no reason, because there arent any apps). For the 4xxx series APU I however need https://github.com/RadeonOpenCompute...ent-1042464940 to make it work, but after this it works so well that content creation is now finally a thing on Linux desktops. Without any form of GPU reset, application crash and everything I suffered since 2015, when AMDGPU replaced FGLRX came up.

                Comment


                • #28
                  Originally posted by You- View Post
                  The issue with their method isnt just whether you need to add a PPA etc. It is also that until now (or maybe still), you were forced to use legacy software versions and if you tried newer software that broke, you wouldnt blame the dodgy driver, but free software.
                  I haven't noticed that. Maybe it doesn't get along with compositing, but I wouldn't know because I've always kept it turned off for the same reason I don't use VSync in games. It pretty much unavoidably increases perceived GUI latency on the non-FreeSync monitors I get as hand-me-downs or buy for $15-30 apiece at the thrift store.

                  Comment


                  • #29
                    Originally posted by horizonbrave View Post
                    Easy: just make sure you don't need CUDA or find an alternative that doesn't require that shit
                    Yeah, like saying to just use an open-hardware video card because nobody really needs 3D acceleration.

                    For example, do you know why my brother runs nVidia on his machine? Because he works in 3D modelling and the non-negotiable tools he needs for his job don't support AMD well enough for him to stake his livelihood on it. (Primarily Blender, funny enough)

                    Comment


                    • #30
                      Originally posted by Aeder View Post

                      There's nothing stopping anyone from porting the old drivers to this modular architecture though?
                      Technically no but realistically it doesn't go down too well. Older pre-GCN GPU's never got AMDGPU, let alone RadeonSI. R600 isn't capable of Vulkan but somehow Raspberry Pi's are. I see this as AMD divorcing away from GCN support.

                      Comment

                      Working...
                      X