Announcement

Collapse
No announcement yet.

NVIDIA Doesn't Expect To Have Linux 5.9 Driver Support For Another Month

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

  • #51
    Originally posted by birdie View Post

    Cancerous ideas keep the world running, you know. You cannot image how many products and services have been designed and created using closed source. Vastly more than what Linux and Open Source have produced.
    Only in your stupid head. Thankfully, Linux proved to be superior in comparison to closed source broken mess and even m$ can't live without Linux. It was noticed many times you're a freak.

    Comment


    • #52
      Originally posted by sdack View Post
      but we still carry around the GPL with its almost militant stance against closed source, which keeps getting in the way.
      I totally agree with you, we should stop this non-sense freedom GPL open source Linux bullshit. All militants of all parties should finally give up they're hopes and ideals and just accept to swallow the truth, NVIDIA is the best, and able to run videogames is the final goal. Fuck you Linus, long life NVIDIA!

      Comment


      • #53
        Originally posted by tildearrow View Post

        Yes, if you are talking about laptops (especially performance laptops).

        Yes, if you are talking about CUDA.
        CUDA? It's made by the devil itself according to Phoronix anal-ysts. CUDA is proprietary crap which must die! Everyone must use OpenCL/Vulkan Compute!

        Laptops? There are hundreds of Intel/AMD iGPU laptops. To be honest I don't understand people who buy laptops with discrete graphics. Switchable graphics only works under Windows and it's not even perfect there. It's a constant never-ending torture.

        Originally posted by jacob View Post

        You know that both can be true, don't you. It's obviously all about what you do. If you mainly play games, their compatibility with NVIDIA is most often better, that's for sure. If, on the other hand, you have a laptop or you run Wayland, then using NVIDIA is like voluntarily getting extra heavy balls and chains to run a sprint race. Across a minefield.
        You see, I want to have work done that's why I don't use Wayland and I couldn't care less about this brain damaged product which provides a very thin layer of APIs over KMS and it's so thin, we have no decent fully functional window managers for it other than those in Gnome 3 and KDE 5. And Wayland is already 12 years old for fuck's sake.

        Comment


        • #54
          Originally posted by Volta View Post

          He's a moron and his contribution is nearly none. He just copied some bug report at lkml. That's all. Linux is much better on desktop than windows ever was. And no, we don't have to agree. You're using semi broken KDE with nvidia proprietary crap, so your experience is far from optimal.
          He's not a moron.

          I'm not using "semi broken KDE". I am using proper KDE and even on the default setup it fails to serve:

          1. Crash, crash, crash, crash every time you do anything in Plasma 5.20. It is fixed now but still... Amazing KDE Quality.
          2. Kdenlive was one of the most unstable video editors ever, to the point I got to crash it in 4 seconds. It was capable of corrupting your projects as well. They fixed this later, but it still can be dog slow and eat your memory.
          3. KWin has stuttering and latency by default, and you need to use something else like my fork or KWinFT to get a smooth desktop.
          4. KWin is broken when using the Wayland session. It is much more stable now but still.
          5. KWin used to crash when doing something as simple as switching windows. Amazing KDE Quality.
          6. Plasma has a chance of surviving the logout and leaving you with a broken login screen.
          7. No joke, the entire desktop CRASHES to logout. Yes. I have set the core_pattern to make a sound when something crashes, and I hear plenty of them upon logout. Amazing KDE Quality.
          8. Sometimes the notification popup wouldn't appear at all, and sometimes it would but it's nothing more than a transparent box that hangs the desktop (which has been fixed as well but still).
          9. Useful features like full-screen unredirection ("exclusive full-screen") not available under stock KWin because they think it is a "hack".
          10. Useful customization options like left-clicking to close windows in Present Windows removed only because the developers "don't feel like having it there".
          11. KDevelop may crash after 4 hours of usage. Amazing. KDE. Quality.

          Furthermore I am not using "NVIDIA proprietary crap". I am using an AMD card, and even so it fails to serve at times:

          1. I am STUCK on Mesa 19.0, because a bug has been introduced in 19.1 which makes your card hang after a while when using FFmpeg/kmsgrab or darmstadt to record the screen. Claimed to be fixed but when I tried out Mesa 19.3.3 the bug STILL WAS THERE! Unacceptable. I want to work. I have NO time to debug the driver.
          2. Even AMDGPU-PRO itself is unstable. My card hung a few times when using it.
          3. The ROCm compute stack is slow, because it wants to "take breaks" when computing.
          4. No proper GPU reset mechanism. If the card goes down for some reason there is no way you can gracefully reset the card without rebooting (losing all your work).
          5. Since Mesa 19.0/Linux 5.0 the GPU clock control is broken, preferring the core clock before the memory clock, which wastes energy.
          6. Since FFmpeg 4.3 it is IMPOSSIBLE to record the screen using VA-API. You are forced to use the .ts or .mp4 format, otherwise it will refuse to record. What kind of disaster is this in where the freaking muxer CARES about the driver?! Just accept the damn H.264/H.265 stream already AND MUX IT TO MKV!! Don't judge me over it!!!
          Last edited by tildearrow; 17 October 2020, 08:28 PM.

          Comment


          • #55
            Originally posted by oiaohm View Post
            That the wrong problem.

            https://nvidia.custhelp.com/app/answ...ploit-advisory.

            Nvidia have published many security bulletins on the Linux/unix drivers. Lot of these like the above one were in fact used in zero days. Gets worse when you know that most of the ones used in zero days if the code had be submitted mainline would have been detected by Linux kernel sparse or one of the automated patch auditing bots.
            Nice URL, so what? I've asked you to show me a single user/company in the world who's been hacked because they were using NVIDIA drivers. Any source for that? No? This URL is a waste of time. AMD has https://www.amd.com/en/corporate/product-security so what?

            Originally posted by oiaohm View Post
            Yes gpu drivers are highly complicated piece of work. You get a point wrong that Nvidia cannot release something that might or might not work. The reality is Nvidia regularly releases drivers for Linux that don't work properly. Worse with fault that would be detected if the code was in fact mainline. Yes I have head of QA/QC the more reviewers you can get to your code base the better. The Linux kernel mainline with the automated bots and tools is a pool of reviewers outside organisation who will test things in different ways
            Proofs about "don't work properly" or GTFO. I've been using NVIDIA GPUs with Linux since the late 90s. I don't recall "don't work properly" ever.

            Originally posted by oiaohm View Post
            I do small work in blender and some cads I have also given up on Nvidia under Linux because the reality is its not reliable enough. Using amd card makes doing my job slower but its reliable this does come from the fact it using a mainline kernel driver.. Other thing is open source drivers feature update lot longer than closed source drivers do so I am able to get a longer productive live out cards in render offloading with AMD over Nvida.
            Small work in Blender in you're already an expert in computing and dare tell the company how to run their business? Have you heard that NVIDIA drivers have been consistently the best in the industry in terms of OpenGL support? No? What other revelations do you have based on your miniscule use case?

            Originally posted by oiaohm View Post
            The horrible part that is also overlooked is the open source nvidia drivers in mesa and the kernel that people attempt to use on desktop cards are partly written by Nvidia so they have open source drivers for their gpu in arm chips. So its really Nvidia being a jackass to themselves here by not releasing the signed firmware so that open source nvidia drivers can work properly on desktop cards to enable broader and more complete testing of their open source driver code base. Yes their open source driver code base its Nvidia made code in the Linux kernel that is not getting as tested as it could be by Nvidia not providing these bits.
            Again, NVIDIA has great drivers which work just fine sans some ass-holeness by kernel developers.

            Originally posted by oiaohm View Post
            Nvidia is shooting QC/QA in foot all over the place.

            Finally the GPL condom in 5.9 kernel if Nvidia is having trouble with it this is a big problem as it makes their driver illegal in china and other countries with equal interpenetration of contract law. China ruling on the matter is very clear. This is technically a big problem if Nvidia does not get it fixed as this kind of fault can prevent Nvidia from using any production FAB in china until they do fix it or cease production of the infringing bit of software. Birdie this should make it clear why you should want a open source mainline Nvidia driver that works. At least if something legal like this happens your means to buy new cards stays possible if properly functional open source driver exists since Nvidia could just pull the closed source driver until it fixed.
            NVIDIA has literally hundreds of millions of GPUs in customers PCs in China. China has supercomputers with NVIDIA GPUs as well. China assembles most NVIDIA GPUs. No idea what you're talking about. Sorry, this whole paragraph sounds like complete mumbo-jumbo for me.

            Again, NVIDIA is a very successful multi-billion corporation and you're talking as if you know for a fact how to allow them to earn even more money. What are you doing here? Go work for Jensen Huang right away. Oh, wait, perhaps your ideas are nonsense and you again believe Linux is the end all be all in the world of computing. Let me tell you a secret: NVIDIA supports Linux as much as it deems required and it works.

            Comment


            • #56
              Originally posted by abott View Post

              You have a completely insane and dumb view of FOSS.

              OSS only needs to care for its self, period. That includes making violators pay for their abuses.

              There's a reason Linux grew so huge and BSD barely lives. Nobody likes being abused by closed source bullshit.
              Angry much? lol. Typical anti-closed source hater. All sound the same. Blah blah blah.

              Chill. Stop making Linux users look angry.
              Last edited by ix900; 17 October 2020, 08:42 PM.

              Comment


              • #57
                I just wonder why there's so much NVIDIA butthurt, hatred and resentment in the topic when Open Source fans can use absolutely perfect Intel and AMD GPUs under Linux whose Linux drivers have zero bugs whatsoever? Why don't you just laugh and move on? Let NVIDIA Linux users have their troubles! Enjoy!

                Isn't NVIDIA the one who's cheating with their drivers as well (this myth has been disspelled many times but anyways)? Bad bad company!

                Comment


                • #58
                  Originally posted by _r00t- View Post
                  RDNA2? What about RDNA1? There's no support yet!
                  This was answered by AMD developers here at the forum much earlier.

                  With RDNA2 RDNA will get a wider use - currently no computational card exist with RDNA, thus ROCm was no priority.
                  Now that the old architectures Polaris and Vega will hopefully be no longer used by future products, the full RDNA stack will come available and get rifinements.
                  So RDNA1 (Navi) will get much better support automatically when this happens (as wiring RDNA2 means wiring RDNA1 first - same for later coming RDNA3).

                  Personally I want to see an RDNA APU capable of 8k with DP 2.0 in near future ... for a real nice silent workstation (maybe after the die shrink to 5 nm - just a dream ). I like my Navi 10 - but coming from a fan-less iGPU system it could be better ...

                  Comment


                  • #59
                    I bought a cheap Nvidia card some years ago, a GT 710, because I just needed something to run a Windows 10 VM I use for music software that doesn't have a Linux equivalent.

                    And to make a long story short, I was shocked when it didn't work because Nvidia is such a lowlife corporation that they attempt to make even home users buy an expensive card, and pay a licensing fee, simply to run a personal VM.

                    I spent quite a bit of time figuring out a work around, but then was shocked again as Nvidia continually worked to assure that home users couldn't use their cards on a VM.

                    So I just put the ridiculous GT 710 in a drawer and spent a little more for a low end AMD card that worked beautifully, and didn't attempt to financially rape me every time I booted my VM.

                    Comment


                    • #60
                      You don't see FreeBSD being dicks making their kernel hard to work with the Nvidia driver. The Nvidia driver works great with FreeBSD.

                      Comment

                      Working...
                      X