Announcement

Collapse
No announcement yet.

AMD's Radeon Gallium3D Starts Posing A Threat To Catalyst

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

  • #61
    Originally posted by KuriKai View Post
    I have the 6870. Using the 3.12 rc kernel. I would also like to add something that benchmarks don't show you, is how the catalyst driver also makes it much better to alt/tab and switch virtual desktop while playing games. Wine games now don't crash when I switch desktop. Dota2/l4d2/tf2 don't freeze while switching virtual desktop.
    Tf2 doesn't have map rendering corruption on the open source drivers.
    I'm getting constant 60fps on the valve games.

    The open source drivers just give a much better experience.
    The is no way in hell I'm going back to the closed source ones.

    Good job AMD and the guys coding the open source driver.
    This issue might be related to your desktop environment or specific hardware... At least on my 7850 with radeonsi in gnome-shell, I can switch workspaces between Eve (running fullscreen in Wine) and another workspace without any issues. So there's hope...

    Comment


    • #62
      Originally posted by BO$$ View Post
      So how exactly are you freer than let's say in the Microsoft business model if everything is done by corporations? You would say that you can fork anytime, but can you actually do that? Can you take a 100 million LOC and jump right in writing code? If the code gets written by a big company, even if it's open source, you still get to do what they tell you since you don't have any manpower. It's just Microsoft with another name.
      You can pay somebody to fix a bug, and in fact, this is often done. Especially after a product is EOL-ed and you still depend on it.

      Now go troll somewhere else.

      Comment


      • #63
        Originally posted by smitty3268 View Post
        The issue doesn't seem to show up in many of the more complicated shader tests - such as Furmark - while being huge in a simple glxgears like test (Triangle) and the less advanced Q3 engine tests. So I'm not sure it's really a matter of optimizing the actual shaders into the compute units, which i would think would tend to give the opposite results. Although who knows, maybe something simple is just spilling 1 extra instruction over the limits and doubling the compute units it takes from 1 to 2.
        Furmark used to be a bit of a special case, since the code was written to load the shader cores as heavily as possible (ie make it easy for the driver pack work into the VLIW cores efficiently). Not sure if that is still the case today.
        Test signature

        Comment


        • #64
          Originally posted by GreatEmerald View Post
          I wonder what's up with the HD 6950.

          Pretty awesome results overall.
          Amount of testing and optimization done for the VLIW4 based GPUs. The only GPUs that used VLIW4 where the HD6900 series, the VLIW5 based hardware was still in use in the low end R7 250 and under series and in the Richland APUs.

          Everything now is moving to GCN based GPUs used in the middle end and up HD7000 series.

          Comment


          • #65
            Originally posted by Kostas View Post
            Is there any webpage tracking Radeon's progress on the post-HD6000 GPUs?
            For rough approximation theres always http://wiki.x.org/wiki/RadeonFeature/

            and http://dri.freedesktop.org/wiki/GalliumCompute/
            not to mention http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt and http://wiki.x.org/wiki/GalliumStatus

            Comment


            • #66
              Originally posted by Panix View Post
              Show me where?!? All I read are complaints about the RadeonSI and the 'major' distros don't have it supported or installed via default.

              Also, the joke that AMD supports Linux is still going on?!? LOL! I agree with what you post here - "Just look at Catalyst -- there are *so many* loose ends -- unfinished features, broken options, plenty of stuff that just flat out *doesn't work* -- and from what I can tell there's no desire to actually fix those things." Does anyone here besides the ''AMD' employees' (FOSS ppl) disagree with that?!? No. So, that is the driver segment that actually gets full funding from AMD? But, it's constantly broken by design.

              I only buy used cards. I don't want to reward AMD OR Nvidia. I'm looking at buying a used AMD card to try since my Nvidia card is really ancient but I am anticipating being disappointed. I don't know what the big fuss is with gaming on Linux. UVD and 2D performance rarely is included in the benchmarks. Doesn't anyone use their computer to watch videos or stream video in Linux?!?
              buy a card that can use r600g. you won't be disappointed.

              Comment


              • #67
                Originally posted by Kivada View Post
                Amount of testing and optimization done for the VLIW4 based GPUs. The only GPUs that used VLIW4 where the HD6900 series, the VLIW5 based hardware was still in use in the low end R7 250 and under series and in the Richland APUs.
                Actually Trinity and Richland are VLIW4 as well -- Llano and Ontario were VLIW5. Kabini is GCN.
                Test signature

                Comment


                • #68
                  Originally posted by duby229 View Post
                  buy a card that can use r600g. you won't be disappointed.
                  Depending which it is. If you want to get UVD, an HD5xxx is probably your best bet: HD3xxx and older and the early HD4xxx series don't have supported UVD.
                  If you wanted OpenCL, you would need an HD4xxx or later gpu.

                  Comment


                  • #69
                    Originally posted by BO$$ View Post
                    So how exactly are you freer than let's say in the Microsoft business model if everything is done by corporations? You would say that you can fork anytime, but can you actually do that? Can you take a 100 million LOC and jump right in writing code? If the code gets written by a big company, even if it's open source, you still get to do what they tell you since you don't have any manpower. It's just Microsoft with another name.
                    You could, i don't know, contribute to the upstream project?

                    Crazy, isn't it?

                    Of course, if the people running the project don't take contributions, then there isn't a huge difference. That's exactly why a lot of people don't like the stuff Canonical is doing, for example, or the projects Oracle/Sun have taken over like OpenOffice. But even then, you can still download the source and make your own modifications locally, even if you can't support an entire fork.

                    Really, come on. This is FOSS 101 stuff here. I know you're just trolling, but this is ridiculously stupid.

                    Comment


                    • #70
                      Originally posted by liam View Post
                      Holy $%!7, did anyone catch gkh's comments about window's driver development (it's around 39:35 in the above video)? They apparently don't have a stable internal api either. There goes the annoying criticism of linux that if only it would keep its internal api stable like windows
                      +1 Thank you for catching this, and yes those people have egg on face.

                      Comment

                      Working...
                      X