Announcement

Collapse
No announcement yet.

Intel/NVIDIA/AMD Compete On Linux GPU Driver Performance

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

  • #31
    Originally posted by drag View Post
    A huge source of this problem is the people that continue to recommend catalyst drivers for ATI to improve performance.

    If they would just give up and realize that Catalyst drivers suck, always have sucked, and always going to continue to suck and completely disregard them as a possibility then it would make things massively easier for new Linux users to understand.

    Instead new users come to places like this and are told that if they want good performance for ATI to install the drivers and they have to struggle through the 'Catalyst is shit' learning curve until they give up and use open source drivers, go back to Windows, or just go out and buy a Nvidia card.
    As long as UVD and powermanagement do not work reliably for all chips hiding the Catalyst driver from newbies is not only non-sensical, but may even end in damaged hardware.
    Go figure what is better to keep them using Linux: Explaining that there are two drivers and why or hiding one and possibly damage their hardware?

    Comment


    • #32
      stil no comparision to closed source Intel Win driver

      Michael promised that a month back. huge dissapointment. This must be some kind of politics. MICHAEL IS A POLITIC. well done Michael.

      Comment


      • #33
        Originally posted by brent View Post
        I have not done any formal benchmarks, but L4D2 and Portal perform about similar in Windows vs. Linux w/ current radeon drivers (and R600_DEBUG=sb). If the open drivers are slower, then not by much.
        TF2 crashes for me with R600_DEBUG=sb on HD4850 as soon as I enter the map.

        what GPU do you have and could you check how TF2 behaves?

        Comment


        • #34
          Originally posted by Nille View Post
          Its looks only like this because the Games engines are very very old. Play something from today e.g. Source Based games or Wine and you see the difference is still huge.
          I would love to see the results from the source games as well.

          Comment


          • #35
            Originally posted by apholux View Post
            Michael promised that a month back. huge dissapointment. This must be some kind of politics. MICHAEL IS A POLITIC. well done Michael.
            It's coming in next few days.... When I'm the only one working on this stuff and have a huge backlog of planned tests, plus lots of other stuff going on, I can't always push out immediate tests especially when it involves doing a Windows installation.
            Michael Larabel
            https://www.michaellarabel.com/

            Comment


            • #36
              Originally posted by Michael View Post
              It's coming in next few days.... When I'm the only one working on this stuff and have a huge backlog of planned tests, plus lots of other stuff going on, I can't always push out immediate tests especially when it involves doing a Windows installation.
              Michael,

              Thanks for running these comparisons - it's always really interesting to see how the different GPUs and drivers compare! Especially interesting to see how Haswell stacks up against the card in my gaming machine, on Linux, using GL.

              Out of curiosity, how long did it take you to run all of these tests? That many combinations must have taken a long time... (I'm not sure everyone here appreciates how much work it is...)
              Free Software Developer .:. Mesa and Xorg
              Opinions expressed in these forum posts are my own.

              Comment


              • #37
                What i find interesting is that it appears Michael has switched his standard testing box from Ubuntu to Fedora KDE, at least for the last several tests.

                It'd be nice to get some kind of statement about that. Just a whim to try it out? Looking to switch away from Ubuntu because of Mir (and in particular, all the XMir slowdowns that will be in the next release)? Just so that it's easier to test out RadeonSI hardware? Why KDE instead of Gnome3?

                I like the switch myself, I just find it interesting and i think others would too.

                Comment


                • #38
                  Originally posted by smitty3268 View Post
                  What i find interesting is that it appears Michael has switched his standard testing box from Ubuntu to Fedora KDE, at least for the last several tests.

                  It'd be nice to get some kind of statement about that. Just a whim to try it out? Looking to switch away from Ubuntu because of Mir (and in particular, all the XMir slowdowns that will be in the next release)? Just so that it's easier to test out RadeonSI hardware? Why KDE instead of Gnome3?

                  I like the switch myself, I just find it interesting and i think others would too.
                  Among the reasons:

                  - Yes, deciding about the future and whether there will be as many Ubuntu Linux tests if it's deviating too far in its graphics stack from the "Linux desktop."

                  - Fedora 19 was just released so it's new and fresh.

                  - During each PTS development cycle I go through a range of Linux distributions to ensure they're all properly detected in terms of hardware/software component parsing, utilizing the package management system, and no other problems.

                  - Doing a clean install I have GLAMOR and other HD 7000 series support ready on Mesa 9.2-devel.... This is a low excuse, but does save me a few minutes of time in my setup procedures.

                  - Why KDE over GNOME3 was simply when downloading the ISO I have been having troubles with a new router (the worst router I've ever dealt with for Linux; an upcoming post) and the KDE download finished successfully before the GNOME3 download.

                  - The default Fedora 19 "out of the box" experience is newer than Ubuntu 13.04 so it provides a more sane, stable point of reference since I still firmly believe in Linux needing a good out of the box / default experience.

                  - Personal curiosity to try out other new features of F19.
                  Michael Larabel
                  https://www.michaellarabel.com/

                  Comment


                  • #39
                    Originally posted by Kayden View Post
                    Michael,

                    Thanks for running these comparisons - it's always really interesting to see how the different GPUs and drivers compare! Especially interesting to see how Haswell stacks up against the card in my gaming machine, on Linux, using GL.

                    Out of curiosity, how long did it take you to run all of these tests? That many combinations must have taken a long time... (I'm not sure everyone here appreciates how much work it is...)
                    You're welcome. Feel free to email me any other test requests, as always,

                    Couple days of testing, but fortunately with PTS most of it's automatic so then it's just a matter of changing out graphics cards / drivers and then casually observing one of the screens to make sure there's no major rendering artifacts or other issues.
                    Michael Larabel
                    https://www.michaellarabel.com/

                    Comment


                    • #40
                      Originally posted by Michael View Post
                      You're welcome. Feel free to email me any other test requests, as always,

                      Couple days of testing, but fortunately with PTS most of it's automatic so then it's just a matter of changing out graphics cards / drivers and then casually observing one of the screens to make sure there's no major rendering artifacts or other issues.
                      First of all, Big Thank You for the work you do. Sure, there are lots of things that could be done "better" or differently with the testing, but you actually go and test this stuff. Kudos.

                      As far as recommendations: could you look into testing some Source engine games (Team Fortress 2, Left 4 Dead 2 or Half Life 2)? I don't have anything against the old style FPSes, but the TF2 player base is quite large...

                      And second thing, I'm quite sure much it's further down the line, could you work on making it possible to measure frame latencies and overall jitter? I'm currently playing TF2, and while the average FPS is above 50-60, the stuttering is really noticeable.
                      Last edited by tomato; 09 July 2013, 04:28 AM.

                      Comment

                      Working...
                      X