Announcement

Collapse
No announcement yet.

The Most Popular NVIDIA Linux News + Milestones Of 2020

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

  • The Most Popular NVIDIA Linux News + Milestones Of 2020

    Phoronix: The Most Popular NVIDIA Linux News + Milestones Of 2020

    NVIDIA's RTX 30 "Ampere" launch was quite a success for 2020 along with new Jetson products and more. Meanwhile on the Linux front this year NVIDIA's proprietary driver continued providing same-day support, features roughly at parity to Windows, and little bread crumbs of open-source support so far. But there still are indications of more possible open-source actions to come as well as potentially better Wayland support to look forward to in 2021...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2

    Comment


    • #3
      let's face it, that opensource announcement is either not coming at all, or it will be very underwhelming.

      Comment


      • #4
        Originally posted by yoshi314 View Post
        let's face it, that opensource announcement is either not coming at all, or it will be very underwhelming.
        Or there is 3rd option that somehow takes longer for them to progress or simply their announcment was supposed to be long term one. One could hope it is that.

        Comment


        • #5
          I believe that for the time being (unfortunately) they'll keep the status quo as it is. If they make an open source announcement, it'll be for some libraries, frameworks, whatever in order to gain more developers in their platform, nothing more than that. At the moment they have no reason (at least in their mind) to open up their drivers/documentation/firmware/whatever. I do think though that at some point it'll hit them back, hard.
          Time will tell though.
          For me, NVidia is a no go for sure. Unfortunately my work laptop has one and I already can't wait to replace it with a truly open source friendly one.

          Comment


          • #6
            Their blob is not even decent by the way. Actually it is a piece of crapware. Nvidia's performance in Linux is worse than unacceptable. It is even significantly worse than in windows, which can be considered as an achievement.
            And I have benchmark scores for that on a 2700x cpu and an 2070 gpu.

            On unigine opengl benchmarks fps are very bad for nvidia linux driver:
            OpenBenchmarking.org, Phoronix Test Suite, Linux benchmarking, automated benchmarking, benchmarking results, benchmarking repository, open source benchmarking, benchmarking test profiles


            And basemark vulkan scores are even worse:
            high settings: linux 7630, windows 8941
            medium settings: linux 81051, windows 103289

            Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

            Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

            Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

            Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.


            Now I can raise the middle finger to nvidia with numbers and not just speculations

            Comment


            • #7
              Well 2020 just finished and still waiting what was the big release to open source from nvidia, I know they release some stuff but I thought will be something huge still waiting.

              Comment


              • #8
                Originally posted by marios View Post
                Their blob is not even decent by the way. Actually it is a piece of crapware. Nvidia's performance in Linux is worse than unacceptable. It is even significantly worse than in windows, which can be considered as an achievement.
                And I have benchmark scores for that on a 2700x cpu and an 2070 gpu.

                On unigine opengl benchmarks fps are very bad for nvidia linux driver:
                https://openbenchmarking.org/result/...HA-MERGE649532

                And basemark vulkan scores are even worse:
                high settings: linux 7630, windows 8941
                medium settings: linux 81051, windows 103289

                Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.


                Now I can raise the middle finger to nvidia with numbers and not just speculations
                To me, the biggest problem and the main reason I avoided nvidia and went with AMD is desktop screen tearing (at least on KDE which I tested when I had access to a PC with nvidia gpu for a few days). Vsync just didn't work unless I turned on full composition pipeline which is a terrible stuttery workaround that also causes input lag. This is generally a problem with binary blobs, same story with AMD's old fglrx blob. I know some on this forum and other forums call those who praise AMD drivers and criticize nvidia's "AMD fanboys", "open source fanatics" and other nonsensical name calling. Just because nvidia works great for you, doesn't mean it works great for everyone else. Personally, I don't even care about what license the driver uses. I just choose what works best for me
                Also, it seems nvidia's vulkan driver on Linux tends to have TONS of bugs and regressions. Performance issues in some Windows Vulkan games, crashes, and performance issues with DXVK.

                Comment


                • #9
                  Originally posted by marios View Post
                  Their blob is not even decent by the way. Actually it is a piece of crapware. Nvidia's performance in Linux is worse than unacceptable. It is even significantly worse than in windows, which can be considered as an achievement.
                  And I have benchmark scores for that on a 2700x cpu and an 2070 gpu.

                  On unigine opengl benchmarks fps are very bad for nvidia linux driver:
                  OpenBenchmarking.org, Phoronix Test Suite, Linux benchmarking, automated benchmarking, benchmarking results, benchmarking repository, open source benchmarking, benchmarking test profiles


                  And basemark vulkan scores are even worse:
                  high settings: linux 7630, windows 8941
                  medium settings: linux 81051, windows 103289

                  Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                  Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                  Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.

                  Powerboard 4.0 is an objective ranking site for smartphones, tablets, PCs and notebooks with easy-to-compare functions. The rankings are based on benchmark scores collected from hundreds of thousands of users.


                  Now I can raise the middle finger to nvidia with numbers and not just speculations
                  You're not even making a fair comparison & then complain about lower performance ... great stuff!

                  Here are the two reasons why Linux (or, more precisely: your botched up bastard version of Linux) is slower than Windows:

                  - Windows uses newer driver series than Linux! (460 vs. 455)
                  - You've compiled Linux with out-of-tree custom patches that severly hamper single-core performance, since tasks are being shuffled around like crazy!

                  Next time, try with the standard CFS process scheduler and post your results here!

                  Comment


                  • #10
                    Originally posted by Linuxxx View Post

                    You're not even making a fair comparison & then complain about lower performance ... great stuff!

                    Here are the two reasons why Linux (or, more precisely: your botched up bastard version of Linux) is slower than Windows:

                    - Windows uses newer driver series than Linux! (460 vs. 455)
                    - You've compiled Linux with out-of-tree custom patches that severly hamper single-core performance, since tasks are being shuffled around like crazy!

                    Next time, try with the standard CFS process scheduler and post your results here!
                    Not a fair comparison?
                    What if I tell you I expect the vanilla kernel to have even worse performance? For example neither O3 optimizations nor march=znver1 are available for the vanilla kernel. Also the pf-kernel does not have a custom scheduler, and my build uses CFS, so your argument about the scheduler is invalid.
                    Considering the driver version, it is not my fault but nvidia's. The 460 series is still in beta for linux (or was at the time I ran the benchmark). Now I have a 460 driver (most likely beta but I am not sure) so I can rerun the benchmark.

                    I will give the benchmarks another try (the basemarks/vulkan ones at least, the unigine stuff take too long) with the latest driver and maybe later with a vanila kernel (with a .config borrowed from arch, just to make sure that the problem is not the exotic kernel). If the results do not improve, I hope that it will be enough proof that nvidia's driver is an utter piece of crapware, even to those that do not approve my setup.

                    Comment

                    Working...
                    X