Announcement

Collapse
No announcement yet.

10-Way AMD GPU Comparison For Team Fortress 2 With RadeonSI Mesa 13.1-dev

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

  • 10-Way AMD GPU Comparison For Team Fortress 2 With RadeonSI Mesa 13.1-dev

    Phoronix: 10-Way AMD GPU Comparison For Team Fortress 2 With RadeonSI Mesa 13.1-dev

    In case you didn't hear, last week a nine year old Mesa bug was fixed that ended up causing stability issues for RadeonSI and was one of the reasons Valve's Team Fortress 2 game wasn't running stable on the open-source AMD driver in quite a while. With Mesa Git now running Team Fortress 2 on RadeonSI without any stability problems, here are fresh benchmarks of that game when using Mesa 13.1-dev and Linux 4.9...

    http://www.phoronix.com/scan.php?pag...-Mesa-13.1-Dev

  • #2
    So RX4xx GPUs are more CPU bound than older cards? What could be the explanation?

    Comment


    • #3
      Originally posted by michal_229 View Post
      So RX4xx GPUs are more CPU bound than older cards? What could be the explanation?
      except 470. i doubt it is real

      Comment


      • #4
        Originally posted by pal666 View Post
        except 470. i doubt it is real
        Yeah, it can be noise.

        Comment


        • #5
          Originally posted by michal_229 View Post
          So RX4xx GPUs are more CPU bound than older cards? What could be the explanation?
          Maybe we can say that 4K results looks strange, because it is newer than FullHD

          And according to strangeness, maybe we can conclude that 128bit Bonaire here is best TF2 bang for the buck and power saver

          edit: altough i am not sure that Polaris 11 should be 20% slower than Bonaire... (P11 should actually be sometimes a bit slower and sometimes a bit faster than even 256bit GCN 1.0 cards like 270X/370) at least by just marketing numbers (and ignoring anything else) my expectation would be that 260X shouldn't be faster than 460, basically 460 should be faster than 260 all the time
          Last edited by dungeon; 16 December 2016, 10:06 PM.

          Comment


          • #6
            @Michael:
            Maybe now is the time for bisecting _your_ R9 290?! ;-)
            No, i don't have one and my system doesn't offer the needed power (Watt) for such an 'old' beast. Nice numbers.

            Comment


            • #7
              So a 7950 performs almost at the same level as a R9 Fury? ugh...

              Comment


              • #8
                Originally posted by theriddick View Post
                So a 7950 performs almost at the same level as a R9 Fury? ugh...
                TF2's entirely CPU-bound. The 7950 is overkill.

                Comment


                • #9
                  Originally posted by theriddick View Post
                  So a 7950 performs almost at the same level as a R9 Fury? ugh...
                  Ideally mass multiplayer games should perform the same on iGPUs and on Titan X, so that anybody can't cheat because of speed

                  Those games are usually blobs even when they are free and recommend blob drivers only, ban wine, etc... ban any too much diversity by design. And all for the reason not because they don't like opensource, but because that prevents more possible cheaters
                  Last edited by dungeon; 16 December 2016, 11:30 PM.

                  Comment


                  • #10
                    Originally posted by pal666 View Post
                    except 470. i doubt it is real
                    Remember that the 470 is overclocked, these numbers do all line up (for what I see) as TF2's use case. some games might benefit from faster access to data, some might need more bandwidth.

                    Comment

                    Working...
                    X