Announcement

Collapse
No announcement yet.

AMD Launches Pitcairn GPUs, Open-Source Not There

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

  • AMD Launches Pitcairn GPUs, Open-Source Not There

    Phoronix: AMD Launches Pitcairn GPUs, Open-Source Not There

    Yesterday AMD officially launched the Radeon HD 7800 "Pitcairn" series as the latest hardware in their Southern Islands family to reside between the Radeon HD 7700 series and their flagship Radeon HD 7900 cards. Unfortunately, the open-source support for these latest AMD GPUs remains unavailable...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    You could have just titled the article something about getting a HD 7950 and you probably have got more hits... LOL

    Comment


    • #3
      Catalyst 12.2 is coming out today I think and Catalyst 12.3 is only 15 days away, so take that into consideration when doing tests.

      Comment


      • #4
        Lack of open-source driver is basically what is keeping me back from buying a new video card. I want it to be the the most durable possible, so I won't buy an older generation one, also I want to but AMD to thank for their openness. And no, Catalyst in not an option thanks.
        Oh well, I was ready to cash out for a 7770 (or even a 7850 now that I saw the reviews) but that'll have to wait. In the meantime Nvidia's Kepler might come out, forcing AMD to reduce the prices. My win, but not AMD's.

        Comment


        • #5
          what is the actual reason, why AMD isnt going intel's way with only providing open source? or maybe rather: how can Intel go open source only, do they have less patents or something?
          i mean, if they are considering using the gallium arch for win8 on mobiles (or whatever the exact statement was), why not dropping catalyst? advertising on the boxes of their hardware that they support linux.
          (if you dont advertise linux its no wonder nobody uses it. imagine how the linux world would grow when someone like amd advertised that they support linux on their product. there is no market? then make one!)

          and something off topic that just came to my mind:
          why cant you reuse the power saving features of the cpu on the gpu?
          (i guess its some sort of algorythm counting how many cycles the proc went idle, compared with the framerate/task to do. -->many idle cycles: ramp down, not fulfilling task/not having finished job, more power!
          but i guess its the timing when reclocking and stuff should occurr, still why is the problem not there for cpus?)

          Comment


          • #6
            Originally posted by jakubo View Post
            what is the actual reason, why AMD isnt going intel's way with only providing open source? or maybe rather: how can Intel go open source only, do they have less patents or something?
            We make 3D workstation products which need very high performance 3D drivers, to the extent that the only practical way to provide that performance for the Linux market is to share code with other OSes. Sharing code exposes enough other stuff that the resulting driver needs to be largely proprietary/binary.

            The workstation business is what funds most of the fglrx (FireGL aka workstation -> fgl) development, so that is where many of the priorities come from as well.

            Originally posted by jakubo View Post
            i mean, if they are considering using the gallium arch for win8 on mobiles (or whatever the exact statement was), why not dropping catalyst?
            Not Win8 on mobiles - Windows Embedded Compact, where having easy access to source code is important. Win8 on mobiles uses the regular Catalyst driver.

            Originally posted by jakubo View Post
            advertising on the boxes of their hardware that they support linux. (if you dont advertise linux its no wonder nobody uses it. imagine how the linux world would grow when someone like amd advertised that they support linux on their product. there is no market? then make one!)
            I don't understand the question/statement.

            Originally posted by jakubo View Post
            and something off topic that just came to my mind: why cant you reuse the power saving features of the cpu on the gpu?
            Do you mean "why can't the GPU power management hardware be designed to be programmed the same way as CPU power management ?" CPU power management is based around having the OS control power state for each core - works OK with 4 or 8 cores but not so good with hundreds or thousands. The two power management models developed independently (in different companies ) -- they are gradually moving together but both will need to change.

            Originally posted by jakubo View Post
            (i guess its some sort of algorythm counting how many cycles the proc went idle, compared with the framerate/task to do. -->many idle cycles: ramp down, not fulfilling task/not having finished job, more power! but i guess its the timing when reclocking and stuff should occurr, still why is the problem not there for cpus?)
            The main short term problem is that GPUs have display outputs, so you can't re-clock the memory while the display is running -- only in a blanking period which is too short for the current PLL programming code to complete. If you don't change the memory clocks you can't change the voltage, and if you don't change the voltage there's not much benefit from changing the engine clock.
            Last edited by bridgman; 06 March 2012, 08:26 PM.
            Test signature

            Comment


            • #7
              ok, thank you! =)

              Comment


              • #8
                Originally posted by bridgman View Post
                We make 3D workstation products which need very high performance 3D drivers, to the extent that the only practical way to provide that performance for the Linux market is to share code with other OSes. Sharing code exposes enough other stuff that the resulting driver needs to be largely proprietary/binary.

                The workstation business is what funds most of the fglrx (FireGL aka workstation -> fgl) development, so that is where many of the priorities come from as well.

                Not Win8 on mobiles - Windows Embedded Compact, where having easy access to source code is important. Win8 on mobiles uses the regular Catalyst driver.

                I don't understand the question/statement.

                Do you mean "why can't the GPU power management hardware be designed to be programmed the same way as CPU power management ?" CPU power management is based around having the OS control power state for each core - works OK with 4 or 8 cores but not so good with hundreds or thousands. The two power management models developed independently (in different companies ) -- they are gradually moving together but both will need to change.

                The main short term problem is that GPUs have display outputs, so you can't re-clock the memory while the display is running -- only in a blanking period which is too short for the current PLL programming code to complete. If you don't change the memory clocks you can't change the voltage, and if you don't change the voltage there's not much benefit from changing the engine clock.
                Could I take the chance and ask why you are restricted on releasing the specs for interfacing with the hardware you sell? Because, to me it doesn't really seem like that would expose either your software parts or the hardware architecture.
                Just curious.

                Comment


                • #9
                  Originally posted by bridgman View Post
                  Do you mean "why can't the GPU power management hardware be designed to be programmed the same way as CPU power management ?" CPU power management is based around having the OS control power state for each core - works OK with 4 or 8 cores but not so good with hundreds or thousands.
                  Wrong as proven by linux-powered clusters. It has nothing to do with driver - cpu power management is handled by pm-drivers inside the OS. The true only question here is either that driver is opensource or closed source, which you always answer with closed source because you get away with that.

                  Comment


                  • #10
                    Originally posted by jakubo View Post
                    what is the actual reason, why AMD isnt going intel's way with only providing open source? or maybe rather: how can Intel go open source only, do they have less patents or something?
                    They don?t care about your questions or needs.
                    They focus on windows and they have agreements with microsoft to a) hinder linux b) hinder opensource c) keep windows implementation best.
                    They don?t care about you.

                    If you ask this question here, you will be fed with a lot of excuses and reasonings, which make no sense. Because those who wants to create, finds ways. Those who is lazy, finds excuses.

                    Your question can be answered very briefly - use those who you appreciate, use intel.

                    Comment

                    Working...
                    X