Announcement

Collapse
No announcement yet.

Trying Out MESA_NO_ERROR / KHR_no_error Support On Mesa 17.2-dev

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

  • Trying Out MESA_NO_ERROR / KHR_no_error Support On Mesa 17.2-dev

    Phoronix: Trying Out MESA_NO_ERROR / KHR_no_error Support On Mesa 17.2-dev

    Merged last month into Mesa Git and improved since then with follow-up commits has been KHR_no_error support for reducing the overhead of the OpenGL drivers by disabling certain error handling for OpenGL games/applications. This in turn can free up some CPU utilization and possibly lead to power-savings too...

    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
    I am not sure it makes sense to always test improvements reducing the CPU overthead (NO_ERROR, Mesa Threading, RadeonSI threading) on extremly powerful CPUs like Xeon, 7700K or Ryzen.

    Comment


    • #3
      Isn't this basically a benchmark for the quality of the game code or the driver code? I'm not totally certain but it sounds like preventing error handling does mean that you get a better performance when there are enough errors without this prevention so that you have significant lower performance without this option.
      Doesn't that imply that there might be issues with Portal while CS:GO seems to be a quite good implementation?
      Last edited by oooverclocker; 23 May 2017, 04:35 PM.

      Comment


      • #4
        Originally posted by oooverclocker View Post
        Isn't this basically a benchmark for the quality of the game code or the driver code? I'm not totally certain but it sounds like preventing error handling does mean that you get a better performance when there are enough errors without this prevention so that you have significant lower performance without this option.
        Doesn't that imply that there might be issues with Portal while CS:GO seems to be a quite good implementation?
        Not necessarily. I'm creating something in Pyglet and even with the most basic rendering, intentionally disabling debug mode has a significant performance improvement.

        Comment


        • #5
          Originally posted by Linuxhippy View Post
          I am not sure it makes sense to always test improvements reducing the CPU overthead (NO_ERROR, Mesa Threading, RadeonSI threading) on extremly powerful CPUs like Xeon, 7700K or Ryzen.
          You should always look for efficiencies no matter the power of the CPU at the time. Soon enough the software will outpace the CPU and you will appreciate all the little efficiencies you can find.

          Comment


          • #6
            Originally posted by Jumbotron View Post

            You should always look for efficiencies no matter the power of the CPU at the time. Soon enough the software will outpace the CPU and you will appreciate all the little efficiencies you can find.
            I was talking about the way the benchmarks were done, not the improvements itself.
            In GPU limited scenarious it doesn't make sense to publish FPSs values when the goal is to highlight the impact of CPU overhead reduction.

            Comment


            • #7
              These results mirror my experience: no gain in Feral games (but no problems either), and crashes with windows games under Wine :-|

              Comment


              • #8
                Originally posted by oooverclocker View Post
                I'm not totally certain but it sounds like preventing error handling does mean that you get a better performance when there are enough errors without this prevention so that you have significant lower performance without this option.
                no. preventing error handling only makes sense when you know that your game does not have errors

                Comment


                • #9
                  I would like to see this same test again on both 1) a lower end CPU (as LinuxHippy suggested), and 2) A lower end GPU.
                  Just to see if there differences scale or not.

                  Comment


                  • #10
                    I have a Sapphire RX470 like Michael's (but mine is a 8GB model), and I'm having a problem with the fans working at full capacity at all times. But in Windows it works silently as it should. Anybody using a Polaris card is having this issue?

                    Comment

                    Working...
                    X