Announcement

Collapse
No announcement yet.

NVIDIA GeForce RTX 3080 Linux Gaming Performance

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

  • #41
    Originally posted by Slartifartblast View Post

    Even so it's still very large, the Intel drivers barely register by comparison. As discrete cards become ever more complex there are also disadvantages of shipping drivers with the kernel.
    Intel does not support a large list of GPUs over multiple different GPU architectures with one driver, they just support a few gens of their very generic integrated GPU. So of course the register header files will be a lot larger for the AMDGPU driver.

    Comment


    • #42
      Originally posted by birdie View Post
      Or take this bug for example. Multiple, literally millions of systems are affected, and Linux developers say some crap why it can't be solved: Well, TBH, it is probably that there is no way to fix this. The root cause is that ACPI claims some resources that will possibly be used by the ACPI AML code, but the native nct6775 driver also requests the same piece of resource.
      Yes, you guessed right, buggy firmware. So what should they do, include some dirty hack just because someone messed up the firmware?

      Originally posted by birdie View Post
      Or take this bug which took me a good amount of debugging and the help of a number of Linux kernel developers including Linus Torvalds himself. The bug rendered a good amount of laptops dead on boot. No one cared except me. People would have given up on Linux entirely if not me. That's all so fucking disgusting. An Open Source dream. More like an open source hell where no one is responsible for anything at all.
      Yea, Linux comes without any warranty. But, what do you wanna do if Microsoft renders your device broken after a update. Sue them? Of course you could but thats kinda useless if you aren't a US citizen as they just ignore foreign curt ruling.
      So thank you for caring, I care too.

      Comment


      • #43
        Originally posted by Alexmitter View Post

        Intel does not support a large list of GPUs over multiple different GPU architectures with one driver, they just support a few gens of their very generic integrated GPU. So of course the register header files will be a lot larger for the AMDGPU driver.
        Hence the second sentence:

        As discrete cards become ever more complex there are also disadvantages of shipping drivers with the kernel.

        Comment


        • #44
          Originally posted by Slartifartblast View Post

          Even so it's still very large, the Intel drivers barely register by comparison. As discrete cards become ever more complex there are also disadvantages of shipping drivers with the kernel.
          It is the source code that is that big, not the actual end result the compiler produces.
          In my books it does not count as bloat. It is all about the end result.
          The header files are suppose to be meta data to help the compiler do its thing, and when used correctly, should not produce any machine code.

          It's 366k lines of the 2.71 million lines of code that is actual C code.
          Compared to Intel:

          For GPU driver size comparison, the Intel "i915" kernel driver supporting form old i915 graphics through Gen12 / Xe Graphics with Linux 5.9 is at 209k lines of code with another 39.2k lines of comments and 48k blank lines.
          So, it is 366k lines for AMD vs 209k lines for Intel, and it is fair to say AMD supports a lot more hardware.
          Last edited by Raka555; 13 October 2020, 07:23 AM.

          Comment


          • #45
            Originally posted by Bobby Bob View Post

            Then make the open stack better than the closed stack and I will use the open stack. Make AMD GPUs faster than NVIDIA GPUs. Make OpenCL better than CUDA. Have day-1 Mesa driver support for AMD GPUs at launch that's superior to the day-1 NVIDIA driver support for NVIDIA GPUs at launch.
            A few point:

            1. The open stack is BETTER than the closed stack -- see MESA, RADV and ACO (on top of wayland, AMDGPU kernel driver, ect) and how performance improve at every release cycle
            2. OpenCL is an OPEN standard, the only disadvantage is that programs (kernel) need to be compiled the first time (nvcc does it beforehand) -- see perf comparison even within NVidia line (CUDA is horribly closed eco-system - I used it for years and moved away)
            3. OpenCL is essentially being replaced by other standards --e.g.: please see Vulkan Compute and ncnn* (I can run YoloV4-lite at 85+ FPS on a rx480)
            4, NVidia day-1 support is via BETA drivers (WOW!), AMD comes with the release cycle of open source projects (e.g. kernel + Mesa); use a rolling distro if you want to take advantage of it

            Then, if you still prefer Nvidia, please go and vote with your wallet.

            *: see https://github.com/Tencent/ncnn

            Comment


            • #46
              Originally posted by Alexmitter View Post
              Yes, you guessed right, buggy firmware. So what should they do, include some dirty hack just because someone messed up the firmware?
              Buggy hardware has existed since the onset of computing in 1981. For some reasons Windows, Mac OS X and Chrome OS all make it work, Linux kernel developers are arguing why they don't want to properly support/enable it. No one gives a fuck! People want to work with Linux, not listen to this "we implemented it according to the specs" crap.

              Originally posted by Alexmitter View Post
              Yea, Linux comes without any warranty. But, what do you wanna do if Microsoft renders your device broken after a update. Sue them? Of course you could but thats kinda useless if you aren't a US citizen as they just ignore foreign curt ruling.
              So thank you for caring, I care too.
              Well, seriously, fuck off. Over 95% of desktop computers in the world run Windows and Microsoft cannot just shit on people like Linux developers do.

              And how nice of you to ignore the rest of my comment. Again, an Open source idealist, more like a zealot who cannot see beyond his nose. It's all perfect in your Open Source world except it's broken left and right and no one gives a damn.

              Comment


              • #47
                Originally posted by Alexmitter View Post
                Yea, Linux comes without any warranty. But, what do you wanna do if Microsoft renders your device broken after a update. Sue them? Of course you could but thats kinda useless if you aren't a US citizen as they just ignore foreign curt ruling.
                So thank you for caring, I care too.
                There are serious issues with your sentence on warranty.

                If you mean support, you can buy it directly from qualified vendors (btw for the _full_ stack, including RDBMs coming with the distro), e.g.: RedHat
                At a fraction of the price asked by Windows, Oracle, ect, etc

                In fact, companies that deploy Linux (mostly servers) do subscribe for support.
                Desktop?
                If it is for home users, you do not need it (try to call MS for your PC problems -- if those are not actually caused by Windows itself -- virus/malware anyone?)
                For corporate desktops, there are solutions (e.g. RedHat, SUSE, maybe even Ubuntu)

                According to MS EULA (AFAIK) their software essentially comes without "warranty" -- once you agree and install

                Comment


                • #48
                  Originally posted by birdie View Post
                  Well, seriously, fuck off. Over 95% of desktop computers in the world run Windows and Microsoft cannot just shit on people like Linux developers do.
                  What!?!?

                  Have you ever been on a call with MS as private user?
                  You are kidding, right?

                  In my last corporate experience it took 4++ weeks to sort out a Windows Authentication issue (Kerberos), while trying to deny that they support it and attempting to sell us any kind of alternative solution.

                  Comment


                  • #49
                    Originally posted by onlyLinuxLuvUBack View Post

                    Buy a quadro , lol
                    Buying a couple hundred of them for work right now... but not for the house

                    Comment


                    • #50
                      Originally posted by birdie View Post
                      Or take this bug which took me a good amount of debugging and the help of a number of Linux kernel developers including Linus Torvalds himself. The bug rendered a good amount of laptops dead on boot. No one cared except me. People would have given up on Linux entirely if not me. That's all so fucking disgusting. An Open Source dream. More like an open source hell where no one is responsible for anything at all.
                      Wow... I've had some hurdles with a Haswell laptop myself, but nothing of that sort (well, it's possibly related to later refactorings around that, intel_iommu=on makes it boot). That looks painful and scary. Congratulations for remaining patient while describing the problem on Bugzilla, despite everything.

                      Comment

                      Working...
                      X