Announcement

Collapse
No announcement yet.

AMD Makes An Evolutionary Leap In Linux Support

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

  • #71
    Originally posted by Dragonlord View Post
    Crossfire is nice on the paper but from a game developers point of view it's not that blistering. I do not see too much future for those multi-graphic cards except misusing one for physics while the other renders... which is though like having a dedicated physics card which helps more.
    Could you expand on that?

    If the game is GPU limited, then you usually get benefits by using Cossfire (either SW or HW). One problem is that modern high end systems struggle to make the GPU the choke point.

    High resolution, high eyecandy, is where the HW becomes limiting, and usually that is where crossfire works best.

    (In reference to a previous comment, CrossfireX is not across family, but allows sibling products work together http://game.amd.com/us-en/content/im...art_July08.jpg

    Regards,

    Matthew

    Comment


    • #72
      As mentioned by you, the pure pixel pusher strength of the card is usually not limiting ( if your shader is complex you either modify it or try to reduce the number of pixels hit by it ). Physics and occlusion culling as well as all the rest of the engine takes a lot of time. Also while fill rate can be augmented like this switches ( render targets [ a lot in DefRen setups ], textures, states ) stay constant. That said I can see it for higher resolutions where pixel count explodes.

      Comment


      • #73
        Sorry but no. They are not even half of the way towards where nvidia is (and even that is a rather lame target).

        Current fglrx drivers suck, bigtime. Not just performance wise (3d is not everything, for example gtk text drawing is 300% faster on the reverse engineered open source driver {r500 class}) but also stability (latest 8.7 catalyst for example crashes ETQW, all before 8.7 result in bad shaders in Savage 2 for a change).

        The thing is just lame. I wonder how big the dev "team" is but I wouldn't be surprised if it was 1 man job.

        Comment


        • #74
          Originally posted by Almindor View Post
          The thing is just lame. I wonder how big the dev "team" is but I wouldn't be surprised if it was 1 man job.
          Lame now, but hopefully these promises from AMD will yield something comparable and eventually surpassing to nVidia, but that remains to be seen. ^^

          AMD has done pretty well performance-wise though with these latest cards. I'm anxious to see more benchmarks though. I don't know if I could really decide between AMD and nVidia right now if I had to as far as just performance, stability, and features, but you may be right in that nVidia is still above AMD with that.

          The only thing I can say for sure is that AMD is definitely working on closing that gap, and it's showing, so I hope they keep up the good work.

          Comment


          • #75
            Originally posted by Yfrwlf View Post
            Lame now, but hopefully these promises from AMD will yield something comparable and eventually surpassing to nVidia, but that remains to be seen. ^^

            AMD has done pretty well performance-wise though with these latest cards. I'm anxious to see more benchmarks though. I don't know if I could really decide between AMD and nVidia right now if I had to as far as just performance, stability, and features, but you may be right in that nVidia is still above AMD with that.

            The only thing I can say for sure is that AMD is definitely working on closing that gap, and it's showing, so I hope they keep up the good work.

            I'd be optimist too, I mean their release pace, packaging and general outlook looks great now, but until they get their act together and stop adding regressions each release, it's all just a music of distant future for me.

            Overall yes, it's getting better, but I could live without the constant bugs and "I wonder what will break this time" feeling whenever I update their driver.

            Comment


            • #76
              Well, I'm going to continue to support them in their efforts. For the most part, the drivers work for me. With the fix of the Screensaver resume, I'm now left with flickering video in Fusion, and HL2 crashing (and that might not even be ATi)

              Comment


              • #77
                Originally posted by he_the_great View Post
                Well, I'm going to continue to support them in their efforts. For the most part, the drivers work for me. With the fix of the Screensaver resume, I'm now left with flickering video in Fusion, and HL2 crashing (and that might not even be ATi)
                The open source ATI driver with Fusion has been really nice at least...

                It'd be nice if driver installation was easier so you could move to different versions easier if drivers were more easily pluggable, and didn't break after each kernel update. Seriously, the Linux kernel needs to be much more modular, that's just annoying for everyone. Would also be nice to see a way to upgrade it without rebooting so you could finally say Linux is completely reboot-proof. Like, if it was modular enough, then the running pieces and control systems could be hotswapped over to the new kernel, and after it was finished transferring, the old one could shut down, would be awesome. Anything is possible with software. ^^

                Comment


                • #78
                  Originally posted by Almindor View Post
                  <snip> for example gtk text drawing is 300% faster on the reverse engineered open source driver {r500 class}<snip>.
                  I think the question was ATI vs NVidia, not ATI vs ATI. We expect that the open source drivers will pick up new framework features (in this case EXA acceleration making use of glyph cache improvements in X server) before fglrx. Also note that the r5xx work is not reverse engineered -- we have display, 2d and 3d documentation packages out for 5xx and that acceleration code was written by Alex (working for AMD) who has full access to our internal HW info as well.

                  You can't say we suck relative to NVidia because our open source driver does something faster than our closed source driver
                  Test signature

                  Comment


                  • #79
                    Well... you mostly suck currently at delivering a stable driver... and that's said without any intend to insult but stating the status quo.

                    Comment


                    • #80
                      Originally posted by Dragonlord View Post
                      Well... you mostly suck currently at delivering a stable driver... and that's said without any intend to insult but stating the status quo.
                      Haven't had any probs here in a long while, but I haven't used any of the new cards yet, waiting for a good gaming need. I wonder how the open nvidia drivers vs. the open ati drivers perform. I do know nVidia's closed drivers are pretty solid though at least.

                      Comment

                      Working...
                      X