Announcement

Collapse
No announcement yet.

KWin May Drop Support For Catalyst, Vintage GPUs

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

  • #41
    Originally posted by hal2k1 View Post
    OK, so why does it apparently get poor performance compared to other software for Linux?

    Note that I am going on hearsay here, I don't use Wine myself, because I use my computers for computing, not gaming.

    Either someone is unfairly bashing the open source drivers by falsely reporting that they are slow with Wine, or or they are reporting it correctly and Wine is indeed buggy/slow with the open source drivers (compared to native Linux games) but not with fglrx.

    So which is it?
    Let's be honest. The OSS drivers are slow when it comes to gaming. Most people won't notice this in non-WINE situations because there isn't much out there for Linux that actually stresses the video cards. But run the Unigine games, or other heavy GPU apps and you will see a large performance difference between the binary drivers and the mesa ones. The mesa devs won't even deny this - they know full well they have a long way to go to catch up with the proprietary drivers. Most linux native games aren't GPU limited, and so it really doesn't matter that the OSS drivers are slower because it's still way over 60fps and no one can tell the difference.

    Second, Wine was built for a decade based on the nvidia driver. Even though steps have been taken to work on other drivers, it shouldn't be surprising that there are still occasionally bugs/issues that pop up on other drivers that the code wasn't originally designed for. I don't believe this is still a major issue, but it can popup from time to time.

    Comment


    • #42
      Originally posted by hal2k1 View Post
      ... my ATI Mobility Radeon HD 5450 GPU should have about the same performance as your HD 4350. They are on the same tier of the hierarchy chart, as is the top of the Intel range in the Intel HD Graphics 3000. Since my HD 5450 and the Intel HD Graphics 3000 both achieve 60+ fps, so too should your system. Are you running Mesa 8 and KDE 4.8 (as used in Kubuntu Precise Pangolin Alpha)? Apparently this gives a very substantial performance improvement over previous versions of the Linux graphics stack.
      I am using KDE3! With basic compositing in kwin like transparency and shadows, nothing else But I can try to disable that. I don't think Mesa 8 has any effect (but I am using it), as we concluded FF is not using OpenGL. I am on the latest radeon driver. But my X server is quite old (like 1.9), maybe that is the problem. I still think your CPU makes it faster for you (I suppose you also see FF using 100% of your CPU core) as I only have a 3Ghz Phenom II 960.
      What version of Firefox have you tested there?

      Comment


      • #43
        Originally posted by AJenbo View Post
        Power managment does not work as well as it does with fglrx and is rather manual.
        Wine works fine with the open driveres, but it's slower and not only with DX games. Even native games are significanly slower when running the open driver then with the blob. Multiple articles on Phoronix will testify to this so i don't see why we are even debating this unless i'm just feeding a trole.
        This is all very, very dynamic.

        There are no articles yet on Phoronix, as far as I am aware, which report the performance of Mesa 8 and the OpenGL 3 milestone (for Intel and AMD/ATI GPUs) which was released in January this year. This release of the Linux graphics stack has very significant performance improvements over previous releases.

        Then again, there are still other very significant performance improvements for the Radeon open source drivers in the form of 2D tiling support which landed after the release of Mesa 8,

        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


        ... and still other improvement which are yet to be finished in the form of HiZ

        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


        Perhaps by the release of Mesa 8.1 later this year the performance gap between the Radeon open source drivers and fglrx will have all but disappeared.

        Comment


        • #44
          Originally posted by aceman View Post
          I am using KDE3! With basic compositing in kwin like transparency and shadows, nothing else But I can try to disable that. I don't think Mesa 8 has any effect (but I am using it), as we concluded FF is not using OpenGL. I am on the latest radeon driver. But my X server is quite old (like 1.9), maybe that is the problem. I still think your CPU makes it faster for you (I suppose you also see FF using 100% of your CPU core) as I only have a 3Ghz Phenom II 960.
          What version of Firefox have you tested there?
          According to Help ==> Troubleshooting Infromation:

          Name: Firefox
          Version: 11.0
          User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:11.0) Gecko/20100101 Firefox/11.0

          According to the System Monitor, under the KDE SC 4.8 desktop Firefox 11 uses 0% CPU when idle (probably due to rounding error) as I type these very characters in this reply window.

          I can momentarily get it up to 3% CPU by using the mouse wheel to rapidly scroll up and down.
          Last edited by hal2k1; 21 February 2012, 06:03 AM.

          Comment


          • #45
            KDE with compositing is already very slow for the 'older cards',
            and there are people that don't need fancy cubes and wobblies
            and flip windows. For those people there are alternatives. KDE
            is pushing on a tablet front, and they really don't need legacy
            support to drag them down. Only time will show whether leaving
            legacy product behind will have any cost. Perhaps it will. But I
            wish them luck... XFCE is good enough for me now

            Comment


            • #46
              Power Management dynamic? Afaik dynpm doesn't work with more than one screen. Also, it still often produces flickering when changing power states.

              Then, I can definitely say that my HD 6550M uses more power on the "low" profile on the Open Source drivers than fglrx. It's not dramatic, but it's definitely noticeable when the power consumption of a laptop goes from ~ 15 Watt to > 20 Watt.
              Also, on this chip the "mid" profile and the "low" profile seem to be pretty much the same. With the exception that on "low" it seems to occassionally crash.
              The performance on either the "low" or "mid" profile is really low. And by really low I mean that even kwin opengl and opengl es effects with raster graphics system and opengl 2 shader disabled produce very laggy window movement. But that has gotten much better lately...
              Last edited by ChrisXY; 21 February 2012, 06:10 AM.

              Comment


              • #47
                Originally posted by hal2k1 View Post
                According to Help ==> Troubleshooting Infromation:

                Name: Firefox
                Version: 11.0
                User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:11.0) Gecko/20100101 Firefox/11.0

                According to the System Monitor, under the KDE SC 4.8 desktop Firefox 11 uses 0% CPU when idle (probably due to rounding error) as I type these very characters in this reply window.

                I can momentarily get it up to 3% CPU by using the mouse wheel to rapidly scroll up and down.
                I mean how much CPU it uses when running that test.
                But it seems you are using 64bit system, so that may be the difference (I am 32bit for now).

                Comment


                • #48
                  Originally posted by hal2k1 View Post
                  This is all very, very dynamic.
                  For one thing you have to manually switch it from static (high load) to dynamic, to get any dynamic power management at all.
                  The conclusion of the folloing articale is also that Catalyst saves more power then the open driver:
                  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


                  Originally posted by hal2k1 View Post
                  There are no articles yet on Phoronix, as far as I am aware, which report the performance of Mesa 8 and the OpenGL 3 milestone (for Intel and AMD/ATI GPUs) which was released in January this year. This release of the Linux graphics stack has very significant performance improvements over previous releases.
                  Phoronix has several articales on the performance of Mesa 8.0 out already, infact the one I already liked to was a comparason between Mesa 7.11, Catalyst and Mesa 8.0, here only HD 6950 on low setting saw any improvements.

                  Originally posted by hal2k1 View Post
                  Perhaps by the release of Mesa 8.1 later this year the performance gap between the Radeon open source drivers and fglrx will have all but disappeared.
                  Ok so we can agree that the radeon drivers is behind fglrx when it comes to preformance. Besides patches are not in yet and most distroes hasn't yet shiped Mesa 8.0 so i'm not relly taking the recent advances in to account.

                  Comment


                  • #49
                    Originally posted by hal2k1 View Post
                    Perhaps by the release of Mesa 8.1 later this year the performance gap between the Radeon open source drivers and fglrx will have all but disappeared.
                    3D performance parity is unrealistic. I usually hear devs saying 60-75% of Catalyst performance is the target..

                    Comment


                    • #50
                      Originally posted by aceman View Post
                      I mean how much CPU it uses when running that test.
                      But it seems you are using 64bit system, so that may be the difference (I am 32bit for now).
                      Oh, I see. On my system when running the HWACCEL test under KDE SC 4.8 Firefox 11 hovers between 5% and 7% except for one interval when it hit 10%.

                      Comment

                      Working...
                      X