Announcement

Collapse
No announcement yet.

AMD Admits It Has Linux Problems

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

  • #41
    How does AMD work for video?
    Does hardware video acceleration work?

    I want to watch HD porno. If AMD isn't porno-compatible, I'm not buying.


    Originally posted by del_diablo View Post
    The isssue is not even that bad. Its just that the most bleeding edge AMD supports is whatever the latest Ubuntu is. Which is annoying for us who wants to have something a bit newer.
    Actually, I've heard that Ubuntu wanted to use the latest X.org server but chose to hold back, because of AMD not having a driver out yet.
    As a Nvidia user, this pisses me off. AMD's behavior is hurting me.


    Originally posted by Gusar View Post
    Where you been, bro? Nvidia 302.11 beta driver has xrandr support.
    Great!
    Now if could only support kernel mode setting too...

    Comment


    • #42
      Originally posted by uid313 View Post
      I want to watch HD porno. If AMD isn't porno-compatible, I'm not buying.
      LOL

      Originally posted by uid313 View Post
      Actually, I've heard that Ubuntu wanted to use the latest X.org server but chose to hold back, because of AMD not having a driver out yet.
      As a Nvidia user, this pisses me off. AMD's behavior is hurting me.
      Well, you could always not use Ubuntu. This is the reason Arch doesn't package Catalyst. Because contrary to Ubuntu, Arch does not want to hold back X because of a proprietary driver.

      Originally posted by uid313 View Post
      Now if could only support kernel mode setting too...
      Err, we already had this discussion. Nvidia *does* have kernel modesetting. What they don't have is a fbcon driver that would work on top of it. You need to differentiate between the concept and a specific implementation. KMS is one implementation of doing modesetting in the kernel, used by the open drivers. Nvidia has their own.

      Comment


      • #43
        Originally posted by bridgman View Post
        Don't think you will in time... big problem here is that non-developers set unrealistically high expectations (basically claiming that the open source drivers would almost immediately outperform the proprietary drivers) and now that original delusion has been replaced with an equally wrong "open source drivers are going to be slow forever" delusion.

        In practice there has been steady progress over the last few years (look at r5xx as an example) and all indications are that performance will continue to improve on the newer GPUs.
        I completely agree about the unrealistically high expectations from non-developers.

        Another aspect is when said non-developers compare vendors on things such as time between hardware release and usable drivers. Intel, for instance had trouble with the Sandy Bridge launch, but had Ivy Bridge support in good shape before the hardware was on the shelves. They've now got 90-something percent of piglit passing on Haswell, which isn't slated for release until this time next year.

        This is undoubtedly related to the number of developers working on the drivers. AMD's market cap is ~1/29th of Intel's, and while I don't think AMD's development team is proportionally smaller, it certainly would benefit from a lot more man-power.

        Comment


        • #44
          In particular, AMD concedes it has the most problems with OpenCL support on Linux. They attribute their Linux problems to the fact that there's many Linux distributions out there rather than just a single platform like Microsoft Windows.
          Stop kidding. You just support few distributions which aren't so different. If you were focus just on Ubuntu I bet it will be the same, because of stupid policy and lack of manpower.

          Comment


          • #45
            I just don't understand why they keep saying 'windows easy to support, linux hard'. And ontop of that mac osx is 'just as easy as windows' well it isn't often said, but ment.

            This, imo is bullcrap. First off, they can make linux support almost as easy. Pick a main stream desktop 'joe user' distro and work for that, keeping in mind (and getting dev input) to not do really weird distro specific shit. I think they mostly do that already, by using Ubuntu (since those get pre-release drivers). It should foremost work there. That's their base. True, the ideal base would be some LSB/linux from scratch something ueber basic. Hell Gentoo would be a really nice base, but best pick something that's extremly mainstream.

            Linux being linux, should let distro maintainers sort the rest. Yes a burden, but not impossible.


            So why would windows or mac osx be any easier to support to begin with? Granted, api's change much less often and abi's are more stable. But you cannot fool me, and tell me that Mac OSX 10.5.* 10.6.* 10.7.* 10.8.* all can use the exact same files and it miraculously just works. And windows? Come on, it's even worse. Let's forget 9x and 2k (even though 2k should be valid still as it's not that much different from XP). WinXP, WinXP SP1, WinXP SP2, WinXP SP3, WinXP SP3..., Win vista SP..., and The same continues for 7 and later 8. Also here, I'm sure you can't just pop 1 driver into all those architectures and it 'just works'. Probably easier, probably less of an issue, especially when you say 'only works on 7' but even so it's still work and effort be put in.

            Comment


            • #46
              Originally posted by oliver View Post
              And ontop of that mac osx is 'just as easy as windows' well it isn't often said, but ment.
              Serious question: Who is developing the GPU drivers for Mac OS?

              Isn't it Apple themselves?

              http://www.macdirectory.com/componen...job_id,200635/

              Comment


              • #47
                Originally posted by Qaridarium
                Getting rid of the FGLRX would double the open-source team and this is the only point what matters!
                How would it double the open source team? I'm quite sure that the FGLRX people are restricted due to proprietary bits, etc.

                Comment


                • #48
                  Originally posted by entropy View Post
                  Serious question: Who is developing the GPU drivers for Mac OS?

                  Isn't it Apple themselves?

                  http://www.macdirectory.com/componen...job_id,200635/
                  I seriously doubt they do all the development.

                  The Graphics team is responsible for shaping the Mac OS device drivers, and working with developers, internal partners, and vendors to define the GPU solutions
                  .

                  I belive that the vendors to work with, are nvidia, intel and amd. Do you really think they can do all that development work to support all these options? Also, amd has released mac hardware in the past, that didn't come with a mac per default, 'aftermarket' card. No, I don't think so.

                  Apple surely works extremely close with amd/nvidia/intel on their graphics bits but I do not believe they do all their driver development in-house based on 'specs' only.

                  Comment


                  • #49
                    " They attribute their Linux problems to the fact that there's many Linux distributions out there rather than just a single platform like Microsoft Windows. "
                    whaaat? repeated lie wince/winxp/win7/winserver x86/AMD64/EMT64/ARM (don't push on "official 'windows' supported m$ policy" there is mobile/server windows too)

                    google typed "opencl linux example" - guess what - "NVIDIA OpenCL SDK Code Samples" examples, code snippets all nice , and deep sublink to http://developer.amd.com/zones/OpenC...s/default.aspx a lot of marketing pages...
                    installed xorg on some pc yesterday with nvidia driver.. opnecl driver note ..

                    They are late and not too active shame i really loved some amd code releases (like SSEX quickmath library)

                    Comment


                    • #50
                      Originally posted by oliver View Post
                      I seriously doubt they do all the development.

                      I belive that the vendors to work with, are nvidia, intel and amd. Do you really think they can do all that development work to support all these options? Also, amd has released mac hardware in the past, that didn't come with a mac per default, 'aftermarket' card. No, I don't think so.

                      Apple surely works extremely close with amd/nvidia/intel on their graphics bits but I do not believe they do all their driver development in-house based on 'specs' only.
                      My question has not been a rhetorical one.
                      I simply don't know it.

                      Let's assume they make use of the available OpenGL cores (from the windows driver implementations for instance)
                      and just wrap around platform specific code - similar to what NVidia and AMD does for their Linux blobs.
                      Then why is it that they still stick with OpenGL 3.2 in Lion?
                      Yes, there might be reasons for that - other than writing large parts of it on their own - but it's strange, isn't it?

                      Edit: http://www.phoronix.com/scan.php?pag...item&px=OTI0OA ("Mac Graphics Drivers Are Still Troubled", March 25, 2011)

                      Apple largely develops their own GPU drivers for Mac OS X in-house and last year they overhauled their OS X OpenGL stack for NVIDIA and AMD graphics processors. This was done as Valve Software was porting their Source Engine to Mac OS X and their current drivers on Mac OS X 10.6 at the time simply didn't cut it. The Snow Leopard Graphics Update improved the OpenGL performance, fixed various corruption issues, improved the GL Shading Language support, and fixed up or added support for various other OpenGL extensions. Like the Mesa / Gallium3D drivers, Mac OS X too is largely living in a OpenGL 2.1 world at this time.
                      Last edited by entropy; 29 May 2012, 06:31 PM.

                      Comment

                      Working...
                      X