Announcement

Collapse
No announcement yet.

AMD Admits It Has Linux Problems

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

  • #31
    Originally posted by uid313 View Post
    People who transcode will burn in hell.
    Encoding should only be done from a lossless source, never from a lossy source.
    I like it hot. The Point is that no hardware decoder can handle High@L5. In the Most cases you has only 720p and 1080p with bitrates that L3.1/L4.0 can handle without an difference.
    And the next on my wish list is are the 10bit profiles.

    Comment


    • #32
      Originally posted by uid313 View Post
      People who transcode will burn in hell.
      Encoding should only be done from a lossless source, never from a lossy source.
      JPEG upscaling should be illegal in USA!

      Comment


      • #33
        Originally posted by Qaridarium
        REAL Problems for AMD are: the "Catalyst" dropping the catalyst on linux and a pure opensource solution would be the gold way to fix this problem.
        As an FGLRX user I like this idea. Sadly, I doubt it would be possible to transfer/shift developers from FGLRX/Catalyst to Radeon (open source) due to restrictions.

        I'm soon going to be getting a cheap-ish NVIDIA card if the AMD situation doesn't improve. Quite fond of their open source work, but the blob is just horrible (I need something more powerful than the open source driver, sadly). I'd much rather see all of the work go into the open source driver whilst leaving FGLRX for dead, but I doubt that'd be possible due to the reason I mentioned above.

        Comment


        • #34
          Originally posted by uid313 View Post
          Originally posted by Adarion View Post
          OpenCL? Omg, always this OpenCL. It's a nice thing to have but there are more important things right now like proper video acceleration and power management on par with fglrx for all chips as well as recent chip support. Okay, they're talking about fglrx but this beast needs support for recent kernel x.org in time, better 2d and better compatibility (e.g. WM compositors)
          Agree.
          Who gives a shit about OpenGL, just make video acceleration and porno work!
          Well, its a executive problem. And its far to easy to understand. You see, a GPU is faster than a CPU, but is limited by its VRAM. A APU bypasses the memory limitation, and its GPU part is faster than a normal CPU. So basically if they get OpenCL working properly, whenever somebody wants to build a render farm they will get a gigantic amount of APUs due insane processing power. But thats not the reason. The reason is a lot simpler: If they have OpenCL, and somebody wants to hit the worlds #1 super computer dream as usual, AMD can ship their APUs to do that. Which also will net them a lot of cash and field testing and a long term contract. Workstation GPUs are also in the same area, special deals at special prices for massive profit.
          Another reason why OpenCL is important is also due the hype of GPU computing. Basically for every application that gets written for CUDA, AMD loses money on a long time scale. For us normal users its irrelevant.

          The irony is that is that just like how FLOSS and OSS is superior to various binary blobs, a FLOSS amd driver could have provided the same, and with potential costumization for the high end computer buyer. The only trouble is "perceived patent problems" from a lawyers point of view, and that it will still take close to half a year to get the driver up to speed with a large and proper team. Its basically a shot towards their own leg, especially as they can't even mantain a proper OpenCL driver for the Windows platform even. Just getting a functionable FLOSS driver would save them a lot of problems, at the cost of a problem thats irrelevant due barrier of marked entry. New entries stealing stuff? Too expensive to even enter the marked. Competitors already got their own solutions.
          The advantage would more than pay off, especially due free driver architecture due the way Gallium is structured. The only problem is that they would be contributing to a collective driver, and as greed dicators: Jews hoards jewgold.


          Originally posted by uid313 View Post
          I would like they to more quickly come out with a new release to support the new ABI of X.org whenever there is a new X.org release.

          It usually take them week or months.
          Nvidia does this much faster.
          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.

          Originally posted by blackiwid View Post
          If I want somewhat stable gaming-fast linux graphics card I naturaly go for Nvidia, the only point in buying a AMD card for linux is because of the opensource driver..
          Enjoy no randr support bro. Its not a problem if you are just using one monitor, but I have heard the horror stories.

          Comment


          • #35
            One more time... "getting rid of the fglrx driver" doesn't provide enough resources to do the same with the radeon driver.

            The whole raison d'etre for a proprietary driver is to allow the sharing of code and developer effort across the entire PC market (ie across multiple OSes). The fglrx driver has good 3D performance because it shares most of the OpenGL stack (and the associated development funding) with other OSes and therefore has access to a much larger dev team. The development work would still need to happen for the other OSes even if fglrx wasn't benefitting from it.
            Test signature

            Comment


            • #36
              Originally posted by del_diablo View Post
              Enjoy no randr support bro.
              Where you been, bro? Nvidia 302.11 beta driver has xrandr support.

              Comment


              • #37
                Originally posted by bridgman View Post
                One more time... "getting rid of the fglrx driver" doesn't provide enough resources to do the same with the radeon driver.

                The whole raison d'etre for a proprietary driver is to allow the sharing of code and developer effort across the entire PC market (ie across multiple OSes). The fglrx driver has good 3D performance because it shares most of the OpenGL stack (and the associated development funding) with other OSes and therefore has access to a much larger dev team. The development work would still need to happen for the other OSes even if fglrx wasn't benefitting from it.
                Indeed- that's pretty much my understanding on the situation. It's just such a shame that I have to sacrifice a proper desktop experience in order to have acceptable 3d for gaming, etc.

                Comment


                • #38
                  Originally posted by Newfie View Post
                  Indeed- that's pretty much my understanding on the situation. It's just such a shame that I have to sacrifice a proper desktop experience in order to have acceptable 3d for gaming, etc.
                  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.
                  Test signature

                  Comment


                  • #39
                    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.
                    Oh, I completely understand that and I am very happy with the overall progress with the open source drivers. The problem for me is not the open source stuff, but rather the poor quality of FGLRX in terms of tearing, slow scrolling, input delays , etc. I already recommend AMD cards to non-gamer Linux users.

                    Apart from 3d performance the open source driver does almost everything I'd ever need it for. It's just not good enough for my needs in that area. Apparently power management is rather lacking on the high-end, too, but I cannot comment on that as I just have a low end Caicos.

                    I've only been using an AMD card on my main machine for about a year, and it has not been a positive experience. It's just the horrible desktop performance. 3D works great, but the tearing and 2d lag drives me insane. The open source driver is *almost* usable for me, but still not quite there yet.

                    Comment


                    • #40
                      Where's the validation?

                      I test Linux software for a living and I've had a AMD video card ( one way or another ) for the last 4 years. The biggest issue I've noticed with AMD video drivers is there doesn't seem to be any validation being done. Yes, I understand that testing Linux can be a complex daunting task because there are a lot of distributions but most of that software side of testing can be tested with a hyper-visor like KVM or VMWARE.

                      I don't understand why don't script out testing? They obviously don't have a good validation process because half the time the driver doesn't install or uninstall properly. These are very basic tests that can be automated.

                      I understand that AMD probably does have some sort of Linux validation but from what I can tell, it need a ton of improvement.

                      just my two cents.

                      BTW - AMD if you are reading this, I'd be happy to take up a contract if you need help.

                      Thanks

                      Comment

                      Working...
                      X