Announcement

Collapse
No announcement yet.

AMDGPU-PRO 17.10 vs. Linux 4.12 + Mesa 17.2-dev Git

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

  • #11
    Originally posted by sonnet View Post

    actually comparing the results they seem too good to be true.
    Then compare more titles.
    Deus EX
    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

    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


    Dirt rally
    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


    Looking at nvidia compared to win10 i can't say that i find the radeonSI performance unlikely, there are some api differences but i still think there's room for improvements.
    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

    Comment


    • #12
      Originally posted by Nille_kungen View Post

      Then compare more titles.
      Deus EX
      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

      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


      Dirt rally
      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


      Looking at nvidia compared to win10 i can't say that i find the radeonSI performance unlikely, there are some api differences but i still think there's room for improvements.
      https://www.phoronix.com/scan.php?pa...-ubuntu-1080ti
      I didn't compare all the results, only mad max.
      The radeon 470 is way slower than the 980ti in windows (like 40% slower )

      It's not possible to see it being largely beaten by with a wide margin by the mesa drivers.

      Comment


      • #13
        Originally posted by sonnet View Post

        I didn't compare all the results, only mad max.
        The radeon 470 is way slower than the 980ti in windows (like 40% slower )

        It's not possible to see it being largely beaten by with a wide margin by the mesa drivers.
        The last in my post was supposed to indicate that nvidia doesn't perform as well on linux OS as it does with windows OS.
        If Nvidia under performs in Linux then there's room for improvements and AMD might perform better in some titles.
        You can't just compare the difference in windows and the apply them on an other OS using different API.
        Given time both drivers may improve and we might see other results.

        Now i haven't done any tests myself so i can't say if something is wrong with the test or not, but seeing how much work that is being put into AMD drivers i do expect to see these kind of results eventually, it took my by surprise to see these results now.

        There was a lot of gains already with 17.1-dev + Linux 4.11 back in March, before all the latest work.
        http://www.phoronix.com/scan.php?pag...mar-2017&num=1
        Last edited by Nille_kungen; 20 May 2017, 04:42 PM.

        Comment


        • #14
          Originally posted by scorp View Post
          Hi Michael, can I find this results on openbenchmarking.org as one profile, or did you do them locally and separately? I would like to make the same run on my machine to compare.
          Thanks!
          Openbenchmarking.org says this:

          AMDGPU-PRO vs. Mesa 17.2-dev
          Tests of AMD Linux drivers for a future article on Phoronix.com

          This test can be reproduced via the Phoronix Test Suite by running the phoronix-test-suite benchmark 1705191-TR-AMDGPUPRO04 command.

          Comment


          • #15
            Originally posted by droste View Post

            Openbenchmarking.org says this:
            Thank you, I'll give it a try.

            Comment


            • #16
              Sick ! Now SteamOS needs to use mesa as the default and the war has been won too.

              Comment


              • #17
                Absolutely loving mesa. I use Mesa-git and just recompile the latest one every week or so.

                Comment


                • #18
                  Wow - awesome results. With this, the OS-Performance scaled 1:1 with its TFlops compared to the CS-Nvidia blob driver, meaning its parity time! *pops champagne* And TFlops are normally a terrible metric to compare to, as it disregards the entire architecture behind.

                  And this is going on... this is still not the end. I... I... who is cutting onions?

                  Thank you so much Devs! You lay the groundwork to kick off a new behavior of consumers which will also result in a different behavior of game-developers.

                  Comment


                  • #19
                    Originally posted by Qaridarium
                    the best AMD could do to end this Nvidia-Monopoly strategy of broken software by design fixed by the closed source driver is to simple END support for any kind of broken software. This would increase the pressure on the software companies to fix there bugs. and this in the end will break the nvidia strategy of broken software by design fixed by closed source drivers.
                    Except that would make everything perform much worse on AMDs hardware, and drive everyone away to Nvidia. Also, it wouldn't stop Nvidia from optimizing for games.

                    Comment


                    • #20
                      Originally posted by funfunctor View Post
                      Sick ! Now SteamOS needs to use mesa as the default and the war has been won too.
                      They will.
                      That's why they are investing in RadV too... Once Vulkan performance is up to par they can drop amdgpu-pro.

                      Comment

                      Working...
                      X