Announcement

Collapse
No announcement yet.

Does anyone know when OpenSource ATI GPUs power options are fixed?

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

  • Does anyone know when OpenSource ATI GPUs power options are fixed?

    I got 6850HD but cannot use it proper due it's power options. Here we are using world the most pricey electricity and using GPU on default bills me much on 7/24 system.
    If I use it on forced low power profile, than animations became jerky.
    Also dynpm method doesn't working well. Can't change clock as needed and also give blinks when changing clocks( from low power to med/high ).

    AFAIK, for a years no one working on it... Why? I believe there aren't much work left.
    I remember on some interview with ATI, they said disabled tessellation on catalyst due lowering power usage, because no game using it and they "love polar bears" much more than others (nVidia). Do they quit loving polar bears?
    I think proper dynpm controls are more important for laptop users due limited batteries.

    Why no one working on it and why we can't use our GPUs as efficient as Catalyst. Does power controlling on GPU has any IPs like DRM too?

  • #2
    Well, I got the answer somewhere here that the power management code alone in fglrx is bigger than the whole mesa codebase... So it's not easy to duplicate its functionality completely.

    http://www.x.org/wiki/RadeonFeature says it's ready but I definitely have more power drain on radeon than fglrx on HD 6550 mobile. Not much, so it's kind of acceptable, but still noticeable.

    Comment


    • #3
      Yes on idle and forcing low profile, there are 10~20 Watts higher than fglrx for my 6850HD.
      But even if you accept this loss, you needed to adjust your card clocking by hand and need to accept jerky graphics also.
      Because there is no good re-clocking... And re-clocking "flickers" screen.

      I wonder If nVidia and/or Intel has similar power problems with their GPUs...
      Might it's time to leave AMD GPUs...

      Comment


      • #4
        I noticed it depends A LOT on the card, on my HD5870 there was no difference between low and catalyst, also dynpm did work flawlessly (on the 5870 or the 3870, I don't remember which one).
        ## VGA ##
        AMD: X1950XTX, HD3870, HD5870
        Intel: GMA45, HD3000 (Core i5 2500K)

        Comment


        • #5
          Anyway I do agree amd should put more attention on power management, I will NEVER buy a laptop with an amd apu if they aren't going to offer proper power management: it's VERY, VERY important on a laptop.
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #6
            Originally posted by darkbasic View Post
            Anyway I do agree amd should put more attention on power management, I will NEVER buy a laptop with an amd apu if they aren't going to offer proper power management: it's VERY, VERY important on a laptop.
            Yep, that is right but: what do you intend to buy then? None of the others is any better.
            In my case I used fglrx til the free driver set was mature enough. (3300 onboard/3870 and now 5670. I actually bought that 5670 at a time where also 6xxx series and first 7xxx were available but it was fine to replace my 3870, use less power and it was better supported at that time by the free drivers. And a bit cheaper. )

            On topic:
            Some years before I measured power uptake of my main system (directly at the wall). And with fglrx it was lowest, then Catalyst on Windows and then free drivers. It was on idle, card was a dedicated 3870. With onboard chips difference was not really much but then they have less functions.
            Today, with power-profiles and "low" setting I got fairly good values on my 5670, comparable to Catalyst on XP. Still it might vary from HW to HW, either by chipset generation or even by card manufacturer (e.g. overclocked or undervolted by default).

            As I was informed by John Bridgman PM is currently a lot of "works everywhere stuff". There are many things that work on a lot but not all HW so they are not activated yet (if I am allowed to quote: "works on many systems but not enough that we recommend using it" ). PM in the Radeon Feature Matrix is relatively generic, maybe it will later be split up in several sub-sections. I followed the development for a long time and it got better and better, sometimes slowly but steady. Of course it's not a matter of a few days so I got used to be patient even though I would love to have a SI series card with working UVD, OGL 4.2 and perfect power management right now.

            Comment


            • #7
              I am not agree. At least my one year old HD6850 on fglrx fails gnome-shell, forcing me log out, couple of time a day...
              Also gnome-shell looks like restarts itself ~ every hour due fglrx... Using 12.3. Don't think if I have same experience with Intel or nVidia cards.

              Also don't research well but don't think if nVidia or Intel has similar power problem on their open source drivers.

              Using GPU with forced "low" power-profile setting is something unhappy.
              Why do we pay hundreds of $ for such a cards if it cannot adjust itself power profile and have jerky desktop animations?

              For example, AFAIK AMD working on open source UVD related decode support, right?
              Instead of working on open source UVD, having proper power management is more important. But no one working on it.
              We can decode movies with our multicore CPUs too. Meaning of having UVD support is lowering CPU usage and have longer battery life as a result...
              But what is meaning of having UVD, if we don't have proper power code? We don't watch movies 7/24 but most of use their PCs 7/24.

              Comment


              • #8
                Originally posted by Death Knight View Post
                We can decode movies with our multicore CPUs too.
                At first there is APU E-350/450 (laptops, nettops and HTPC) that can't decode 1080p video itself (on CPU) but have UVD for that. Second - there is video files that not every dual-core CPU may decode - 1080p Hi10P video requere much more resources for deconding. I hope UVD support in R600g will solve this two problems.
                Originally posted by Death Knight View Post
                Meaning of having UVD support is lowering CPU usage and have longer battery life as a result...
                But what is meaning of having UVD, if we don't have proper power code?
                Proper power management is big and difficult task. Probably like GPGPU or maybe more difficult, because mistake in power management cost too much for end-users (check long Intel RC6 story - now it's enabled by default, but there is still RC6-related issues).

                Comment


                • #9
                  I think AMD APUs are better to stay with fglrx. Open source drivers has power issues, right? APU devices are generally battery depended devices and power saving is more important than UVD decode (I believe this is true for the most).
                  Also don't know about h264 Hi10P and it's computational requirements but dual cores could decode most of the movie, which has low movement on scene. For high bitrate scenes, what is "frame drop" option for?

                  Yes watching movie without UVD will drain battery at the other hand... But I don't believe if majority of APU owners dying for watching movie on their portable device, on battery. For people using HTPC, they are "plugged" and if have suffer from mpeg decode, than could continue to use with fglrx.Why they needed to use open source?

                  Working on power options might be hard. But I believe designing a C/G/APU is more harder task and AMD doing it already, right?
                  I think all they need is hire one or more engineer (if required) for the problem and spend some money on task...

                  Comment


                  • #10
                    Originally posted by Death Knight View Post
                    I don't believe if majority of APU owners dying for watching movie on their portable device, on battery.
                    APU devices are generally battery depended devices
                    Not always - nettops and HTPC
                    Originally posted by Death Knight View Post
                    For high bitrate scenes, what is "frame drop" option for?
                    For high bitrate scenes there is UVD.
                    Originally posted by Death Knight View Post
                    For people using HTPC, they are "plugged" and if have suffer from mpeg decode, than could continue to use with fglrx.Why they needed to use open source?
                    Because UVD support in fglrx is very limited in compary with Intel driver or nVidia driver or Catalyst for Windows.
                    Originally posted by Death Knight View Post
                    Working on power options might be hard. But I believe designing a C/G/APU is more harder task and AMD doing it already, right?
                    Yes, so what?
                    Originally posted by Death Knight View Post
                    I think all they need is hire one or more engineer (if required) for the problem and spend some money on task...
                    They already do that.

                    Comment

                    Working...
                    X