No announcement yet.

Linux 3.5 Can Massively Boost AMD Radeon Graphics

  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux 3.5 Can Massively Boost AMD Radeon Graphics

    Phoronix: Linux 3.5 Can Massively Boost AMD Radeon Graphics

    The Linux 3.5 kernel is capable of delivering some massive performance gains for some of the more recent generations of ATI/AMD Radeon graphics processors. Here's some benchmarks showing the hefty performance gains found when using the latest kernel that is still being developed.

  • #2
    That's impressive. Here I thought that only small improvements are left to gain, but no! I am happy for this development. Keep up the good work, guys. I also hope for open drivers to follow the path of workstation quality drivers. I mean "no lossy optimizations".


    • #3
      All the bad news about AMD Radeons, and now comes the strawberries and cream.

      I've always purchased Radeons, but I sometimes play Everquest and the graphics messes up with this line of gfx cards. I was thinking of going Nvidia for a season. What will sway me is the price of the GFX660 and AMD's public relations and driver quality. I will be upgrading when GFX660 comes out and it will vs the HD7870 or possibly the HD7950 if the prices have shifted. Looking to spend about $280-350AUD.


      • #4
        Forgot to say...

        Finally the work pays off for Linux, with the last few years seeing massive changes in GFX rendering direction. It can only get better!

        Thanks to all the coders !!!


        • #5
          With all the talk about Diablo3, I pulled out my NWN1 CDs and installed it as wine version because I couldn't get the Linux installer to work (uses OpenGL 1.2 ). Works fine on my NI radeon.

          Now I am addicted again and in love with Aribeth de Tylmarande...


          • #6
            See to believe

            Well, in opposition to the hype I've seen in the article, I'll take a more "conservative" approach: when 3.5 stable arrives, I'll give a new try to the FOSS ATI drivers (I went back to catalyst because of VAAPI and proper PM). From what I've seen, we'll have better 3D performance from now on (but does that apply to r600/r700 cards? I didn't see any card of that generation tested in the article...).

            Furthermore, I'm also expecting that the "default" profile will stop using always the maximum frequency of the GPU, because that kills the GPU lifetime. (Nouveau does the opposite, btw...)

            Finally, in the medium/long run, It'd be nice to have H.264 VDPAU/VAAPI/UVD acceleration. That wold be nice for those who still have weak CPUs (e.g. AMD E-350/Low-End Llanos/Nehalmen Core2Duos).



            • #7
              Really impressive stuff . Even more if on all those tests the "tweaks" were disabled. Congrats to the developers.
              Last edited by 123_qwe; 06-07-2012, 09:27 AM.


              • #8
                I <3 radeon devs. Thanks so much for your great efforts! With these backendmap fixes, we will be seeing hi-z merged soon! Thanks Jerome!

                This is a great time to have an AMD graphics card on linux!


                • #9
                  Hmmm, Now I'm interested

                  When can we expect to see some results where the OS graphics drivers can compete with the FGLRX binaries?
                  Also, we can we expect some benchmark comparisons? I'd love to switch to the OS drivers on my AMD ASUS 6950 DC II .
                  Can and will it ever support eyefinity?

                  These are my top questions. ;D


                  • #10
                    Michael publishes radeon vs Catalyst benchmarks periodically; IIRC the last one was in the last month or two.

                    Here's the last one :


                    AFAIK the core Eyefinity functionality (lots of screens working together) is available today in the open drivers. Some of the proprietary features, eg automated bezel compensation and convincing the games to keep their HUD on one screen) are not available, although I think you can use randr functions for bezel compensation.
                    Last edited by bridgman; 06-07-2012, 10:42 AM.


                    • #11
                      This is a really impressive performance increase. Still not enough to compete with Catalyst, but more than just noteworthy. I'm curious to know if this affects significantly older generations of GPUs, like the pre-HD2000 series cards.

                      It's too bad Debian sid doesn't like to stray from the 3.2 kernel, I guess I won't be experiencing this for a while.


                      • #12
                        So what's the bottleneck for the 6450? Call it "crippled" all you like, those results are anomalous and should probably be investigated.


                        • #13
                          Alex would know for sure, but my guess is that it only has a single memory channel so improving the way accesses are mapped across memory channels doesn't help.


                          • #14
                            The issue that was fixed only affected cards with >2 render backends (rv740, cypress, juniper, barts). Other chips were not affected. Parts like 5450 and 6540 are generally memory bandwidth limited as they have fewer memory channels and often only have ddr memory rather than gddr memory.


                            • #15
                              Originally posted by Wyatt View Post
                              So what's the bottleneck for the 6450? Call it "crippled" all you like, those results are anomalous and should probably be investigated.
                              Not really if you read the article, or patches.

                              The fix is only for certain GPUs which were misconfigured previously, rv740 is the only r600/r700 card, and then a bunch of evergreen/ni cards, but only in the higher end GPUs.

                              Its not meant to speed up all cards, its more that these cards were broken before now.