Announcement

Collapse
No announcement yet.

Enabling HyperZ Is Still An Easy Way For Faster RadeonSI Performance

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

  • Enabling HyperZ Is Still An Easy Way For Faster RadeonSI Performance

    Phoronix: Enabling HyperZ Is Still An Easy Way For Faster RadeonSI Performance

    With this week's Catalyst vs. RadeonSI Gallium3D driver comparison for the A10-7800 Kaveri APU with Radeon R7 Graphics, the default driver settings were used since after all it's what most Linux gamers and desktop users will utilize when running either driver. However, for those wanting to tune their open-source driver to get a bit higher number, enabling HyperZ is still an easy win.

    http://www.phoronix.com/vr.php?view=20768

  • #2
    did anyone check for performance vs ondemand govenor on catalyst and radeon lately? or is the issue completely done for?

    Comment


    • #3
      so much talk

      so much talk about amd but they contnued without do some good support for their open driver

      Comment


      • #4
        Originally posted by jakubo View Post
        did anyone check for performance vs ondemand govenor on catalyst and radeon lately? or is the issue completely done for?
        I use performance, with ondemand there is overall minus 7-12% difference for me on Kabini (with much sttutering in some games). That on let say OpenArena/Xonotic benchmarks.

        But also, hyperz on and off is also night and day here on Kabini . Hyperz saves bandwith, so on lower bandwith machines seems like gap is bigger so for me that difference in OpenArena/Xonotic is nearly +40% difference that is why i want to have this feature no matter what .

        If these Micheal results for Kaveri are right hyperz is fine to have of course, but i am not impressed there .

        Comment


        • #5
          What is the performance per watt like with HyperZ?

          Comment


          • #6
            I just did, the ondemand issue is still present for AMD FX procs

            Originally posted by jakubo View Post
            did anyone check for performance vs ondemand govenor on catalyst and radeon lately? or is the issue completely done for?
            I just ran a quick test in Critter, a 2D game written in OpenGL that generates very high framerates and may not be representative, but I am still getting at least
            20% more frames per second with the CPU set on full speed instead of on "ondemand." For some reason video rendering also benefits from being set to full speed
            even though you'd think such a heavy load would keep it pinned there anyway,

            Using AMD FX 8120 overclocked to 4.4GHZ with a Radeon HD6750 in both tests.

            Comment


            • #7
              Sounds similar to the Triangle test (which also ran very slowly in the benchmarks). Hmmm...

              Comment


              • #8
                Originally posted by bridgman View Post
                Sounds similar to the Triangle test (which also ran very slowly in the benchmarks). Hmmm...
                Kabini is faster here in Triangle test then Kaveri . I have (with radeon driver of course):

                12359 Points with nohyperz
                21773 Points with hyperz



                That is 43% diff, while for Kaveri diff is 24%.

                Summa summarum my Kabini is 17% faster then Kaveri with hyperz enabled, but 9% slower without .

                But better try to reproduce this test, maybe somewhat is wrong with Michael's numbers .

                Comment


                • #9
                  Originally posted by Luke View Post
                  I just ran a quick test in Critter, a 2D game written in OpenGL that generates very high framerates and may not be representative, but I am still getting at least
                  20% more frames per second with the CPU set on full speed instead of on "ondemand." For some reason video rendering also benefits from being set to full speed
                  even though you'd think such a heavy load would keep it pinned there anyway,

                  Using AMD FX 8120 overclocked to 4.4GHZ with a Radeon HD6750 in both tests.
                  Just checked, so yeah 14% difference for me in Critter . Only i needed some time to figure out that Critical Mass is Critter, or maybe i am wrong... you mean this Critter?

                  http://criticalmass.sourceforge.net/

                  Because there is some different Critical Mass game on Steam too .

                  Comment


                  • #10
                    Yes, Critter is Critical Mass

                    Originally posted by dungeon View Post
                    Just checked, so yeah 14% difference for me in Critter . Only i needed some time to figure out that Critical Mass is Critter, or maybe i am wrong... you mean this Critter?

                    http://criticalmass.sourceforge.net/

                    Because there is some different Critical Mass game on Steam too .
                    Sorry about that, for some reason I know the game by the name of its executable. Yes, they are the same, and this game's an old favorite of mine given I enjoyed the games of the early '80's. Can't find ports of Paperboy, Crossbow, and Centipede though...

                    Comment

                    Working...
                    X