Announcement

Collapse
No announcement yet.

NVIDIA vs. AMDGPU-PRO vs. RadeonSI OpenGL Comparison For Holidays 2016

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

  • #31
    Originally posted by bug77 View Post
    You either have a funny definition of "really good" and "amazing" or you can't see the green bars
    you can't see nouveau bars because nouveau wasn't tested.

    Comment


    • #32
      Originally posted by Dea1993 View Post
      it's a pity that mobile APU performance are really bad, seems that they are optimizing only for desktop GPU and not for mobile hardware
      mobile apus just don't have $612.00 xeon cpu which was used in this benchmark

      Comment


      • #33
        Originally posted by bridgman View Post

        I don't think it's even possible to "optimize only for desktop GPU and not for mobile hardware".

        I believe that essentially all of the optimizations "being done for desktop" would show comparable improvements when running on mobile hardware, even if the result is going from 20 fps to 40 rather than from 40 to 80.

        If you are saying "the numbers are lower on mobile HW" that's probably true - mobile hardware tends to use lower end GPUs (fewer CUs, narrower/slower memory) and both CPU and GPU clocks are lower for power efficiency, but that's about it.
        This actually brings to my mind..

        I have a laptop with "AMD hybrid graphics" iirc. It has a Radeon HD 6520G and a Radeon HD 7670M. On Windows running dota 2 with medium quality settings (with the OpenGL renderer) i achieve a bit over 60 fps. On Linux with the mesa drivers i achieve less than 20 fps, 16 is a pretty average number. It doesn't matter if i do the DRI_PRIME=1 shenanigans, the perf is just as bad with both of those GPU's. Is there any logical explanation for such a big perf difference?

        Also, that hardware combo is "broken" out of the box. I remember how it spilled something about DPM (dynamic power management) having some issues when shutting it down, and upon some googling i came across a bugzilla ticket with a commit that should've disabled DPM on that particular hardware, but interestingly it wasn't disabled for me. Had to manually add the flag for disabling DPM support. After some digging i even found the bugzilla ticket; https://bugzilla.kernel.org/show_bug.cgi?id=51381 comment #38 has the same HW as i do and Alex apparently disabled DPM for it but that definitely wasn't the case for me.

        Oh well, that's what i get for running old hardware i guess.

        Comment


        • #34
          Originally posted by Geopirate View Post
          Nvidia has one driver with different shims for different platforms. This means the bulk of their engineers are working on the same code that works across all platforms
          this also means their driver doesn't really work on linux

          Comment


          • #35
            Originally posted by FishPls View Post
            On Windows running dota 2 with medium quality settings (with the OpenGL renderer) i achieve a bit over 60 fps. On Linux with the mesa drivers i achieve less than 20 fps, 16 is a pretty average number. It doesn't matter if i do the DRI_PRIME=1 shenanigans, the perf is just as bad with both of those GPU's. Is there any logical explanation for such a big perf difference?
            Can't edit my post..

            But i forgot to mention that on Linux, i get the ~16 fps with _low_ quality settings, on medium it's practically a slideshow.

            Comment


            • #36
              And one last thing, i made a mistake in my post; It's not actually DPM support, it's just runpm support as can be seen from the bugzilla ticket

              Comment


              • #37
                Originally posted by FishPls View Post
                Oh well, that's what i get for running old hardware i guess.
                it probably uses crossfire on windows, which is not going to be supported on linux in opengl at least, but that hardware is not going to get vulkan driver either

                Comment


                • #38
                  Originally posted by FishPls View Post

                  Can't edit my post..

                  But i forgot to mention that on Linux, i get the ~16 fps with _low_ quality settings, on medium it's practically a slideshow.
                  Tried gallium nine ?

                  Comment


                  • #39
                    Originally posted by duby229 View Post

                    No clues? Really? Nvidia doesn't care one tiny little bit about specifications.
                    Do people even know what SPECIFICATIONS are or mean? lets me honest, if the standard is telling you to program a shader/code in a way that is going to mean you loose half your performance, then I don't blame NVIDIA trying to optimize on top of standards that are clearly not doing AMD any good by following them...

                    Comment


                    • #40
                      Originally posted by FishPls View Post

                      This actually brings to my mind..

                      I have a laptop with "AMD hybrid graphics" iirc. It has a Radeon HD 6520G and a Radeon HD 7670M. On Windows running dota 2 with medium quality settings (with the OpenGL renderer) i achieve a bit over 60 fps. On Linux with the mesa drivers i achieve less than 20 fps, 16 is a pretty average number. It doesn't matter if i do the DRI_PRIME=1 shenanigans, the perf is just as bad with both of those GPU's. Is there any logical explanation for such a big perf difference?
                      Hybrid graphics doesn't work under linux. You are probably running off of the power efficient GPU all the time with it never using or switching to the more powerful one.

                      Comment

                      Working...
                      X