Announcement

Collapse
No announcement yet.

Marek Boosts Glxgears Performance By 20% For Christmas

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

  • Marek Boosts Glxgears Performance By 20% For Christmas

    Phoronix: Marek Boosts Glxgears Performance By 20% For Christmas

    Well known open-source AMD driver developer Marek Olšák has taken to some Christmas day hacking on Mesa with a significant performance improvement for AMD APU owners and those who care about glxgears...

    http://www.phoronix.com/scan.php?pag...-Glxgears-Xmas

  • #2
    while it's not a benchmark I usually test it for throttling it usually drops a few fps if computer is gpu throttled.
    On laptops it's useful to see if the cpu is throttled or gpu to know which side to tilt for more airflow.

    Merry xmas!

    Comment


    • #3
      Aren't color clears important?

      Comment


      • #4
        It's a Christmas miracle! It was written in the New Testament (Vulkan specifications) as well as in the Old Testament (OpenGL specs) that open source drivers performance will rise from the benchmarks and banish the wreched green dwarf from our cases

        Comment


        • #5
          I'll sit watching glxgears till the end of 2017.

          Comment


          • #6
            Originally posted by oleyska View Post
            while it's not a benchmark I usually test it for throttling it usually drops a few fps if computer is gpu throttled.
            On laptops it's useful to see if the cpu is throttled or gpu to know which side to tilt for more airflow.

            Merry xmas!
            Exactly the thing I do. It's really good for heating up GPU's, just recently MSI "frozr" cooler on my GPU died, while it is of good quality and served it's purpose well, it's a hell to fix it, and if you are not very careful you may damage it and make it unusable, that's exactly what I did lol. Slapped big low RPM cooler on GPU and tested with glxgears. silent and cool , and good thing is that RPM is controled by voltage, so even tho cooler does not have sensor or connections for it, it still goes up when in 3D mode for a while, so it works at even lower RPM normaly.

            Comment


            • #7
              Ok..... Yeah, totally correct glxgears is not a benchmark. I don't honestly think claiming a percentage improvement is even possible. With glxgears is a 20% gain even possible to perceive? I really seriously doubt it considering simply moving the window around varies the framerate by thousands of percent. Having the window in a different spot on screen can vary framerate, I just don't see any possible way to collect data points from it in a way that can result in the conclusion that performance improved in some way.

              EDIT: Literally the only thing glxgears does is draw frames, most of a GPU's hardware doesn't even get touched by it at all. It is -NOT- a good indicator for thermal performance for that reason. Even if you had hundreds of glxgears instances running it still would be completely incapable of stressing any GPU.
              Last edited by duby229; 12-25-2017, 12:39 PM.

              Comment


              • #8
                given the patch is simply changing where the boundary between "no need for fast clears" and "oh, now we need them" - I think there was just a little bit of satire.

                on the plus side - woot, more performance tuning :P

                Comment


                • #9
                  Typo:

                  Originally posted by phoronix View Post
                  this simple chage makes glxgears

                  Comment


                  • #10
                    Originally posted by duby229 View Post
                    Ok..... Yeah, totally correct glxgears is not a benchmark. I don't honestly think claiming a percentage improvement is even possible. With glxgears is a 20% gain even possible to perceive? I really seriously doubt it considering simply moving the window around varies the framerate by thousands of percent. Having the window in a different spot on screen can vary framerate, I just don't see any possible way to collect data points from it in a way that can result in the conclusion that performance improved in some way.

                    EDIT: Literally the only thing glxgears does is draw frames, most of a GPU's hardware doesn't even get touched by it at all. It is -NOT- a good indicator for thermal performance for that reason. Even if you had hundreds of glxgears instances running it still would be completely incapable of stressing any GPU.
                    I don't know, for me, FPS does not change if I move window..., but you are right GPU load is not high when not maximized, but when maximized it goes to 100% for me.

                    Comment

                    Working...
                    X