Announcement

Collapse
No announcement yet.

CS:GO & TF2 Extensively Tested On The Newest Open-Source Radeon Linux Driver

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

  • #21
    Originally posted by stqn View Post
    That resolution is pretty atypical; I’d have liked to see a more usual 1920?1080 resolution.
    As stated in the article, the resolution was chosen to reduce CPU bottleneck. He didn't choose 4K for compatibility reasons. However, some GPUs in this list may have ended up bottlenecked by VRAM. They could otherwise be performing significantly better.


    It could be interesting if Michael were to use his best AMD GPU (which appears to be a R9 290) and using a not-so-graphically-intense test, check the lowest resolution he can use where the CPU isn't a bottleneck. Then, use that resolution for all tests. That way, the only bottleneck GPUs should encounter is with the GPU core itself.

    Or as another idea, he should have a separate test system strictly for GPU tests where he uses a very average CPU (like an i5) and heavily overclock it, so the CPU can't be a bottleneck.

    Comment


    • #22
      @schmidtbag

      See those results from 2 days ago, if i7-5960X at 4GHz is not enough i don't know might be somethin else happens there




      You may try to reproduce those , so is your CPU also bounded with those cases?
      Last edited by dungeon; 12 November 2014, 04:07 PM.

      Comment


      • #23
        If r9 290/x cards drop to $180 - $220 this fall, considering the imminent release of the 390x, and all the Black Friday stuff and potential ebay sales, I'll definitely look to pick one up to play with on radeonsi. Good work AMD dudes!

        Comment


        • #24
          Originally posted by schmidtbag View Post
          Yeah, I often disable compositing for games, when I remember to; I didn't remember to last night. I am 100% sure I'm using radeon, and I usually do disable vsync. I'm going to try disabling KDE's vysnc, disable compositing, and then try playing a game (with vsync) and see how that works.


          You just need to mark the option to disable compositing in fullscreen aplications, in System Settings, nothing more. With this option checked, KDE will perform the same as others DE. I have a R9 290 too and the Vsync options inside the games are independent from KDE options.

          Radeonsi drivers have stuttering in some games, but I am sure that every single game got some jump in fps. I upgraded from a 6970 and have seen some games gaining about 40% in fps.

          Comment


          • #25
            Would be nice to also get minimum FPS, along the average FPS in future. But thank for the current benchmark.

            Comment


            • #26
              Originally posted by iniudan View Post
              Would be nice to also get minimum FPS, along the average FPS in future. But thank for the current benchmark.
              The Source Engine doesn't expose min/max FPS in any useful way AFAIK which is why It isn't picked up by PTS and shown on graphs like it is for some other tests.
              Michael Larabel
              https://www.michaellarabel.com/

              Comment


              • #27
                Originally posted by zanny View Post
                If r9 290/x cards drop to $180 - $220 this fall, considering the imminent release of the 390x, and all the Black Friday stuff and potential ebay sales, I'll definitely look to pick one up to play with on radeonsi. Good work AMD dudes!
                Imminent how? Last I checked, the R9 390X will not emerge on the market until the first half of 2015.

                Comment


                • #28
                  Originally posted by Michael View Post
                  The Source Engine doesn't expose min/max FPS in any useful way AFAIK which is why It isn't picked up by PTS and shown on graphs like it is for some other tests.
                  Ok, thank for the answer.

                  Comment


                  • #29
                    Originally posted by iniudan View Post
                    Would be nice to also get minimum FPS, along the average FPS in future. But thank for the current benchmark.
                    libframetime would allow that for all GL apps no matter what numbers they expose. But...

                    a) someone would have to code the PTS support
                    b) it may still not get used because you need to install it (even if it's packaged in Ubuntu, etc)

                    Comment


                    • #30
                      mesa performance improved a lot in the last week. at least for my HD7770.

                      here my testresults.
                      http://openbenchmarking.org/result/1...KH-TEST1624256

                      i tested distributed kernel+ fglrx 4.10.2 against 3.17.3 mesa from 14th Nov and 21 Nov

                      for fglrx I didn't run the full pts/gaming suite yet. And I was a bit lazy when choosing the names. sorry for that.
                      but the results are impressing.

                      Comment

                      Working...
                      X