Announcement

Collapse
No announcement yet.

ATI Gets Dynamic Power Management & Profiles Too

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

  • #11
    Just checked out the new tree and at least for me it's a huge step backwards.

    It doesn't work at all for me. Fan is spinning like crazy and I don't see any kind of clock reduction at all. Have played around with both power_method and power_profile and neither of them do have any effect on the clocks.

    Watching /sys/kernel/debug/dri/0/radeon_pm_info verifies this, the informations stay static all the time.

    Switching drm.debug=1 doesn't reveal anything interesting, the pm just does nothing for me.

    EDIT: I'm not even getting a overview of the power states like in previous versions of the pm code.

    Comment


    • #12
      Originally posted by LiquidAcid View Post
      It doesn't work at all for me. Fan is spinning like crazy and I don't see any kind of clock reduction at all. Have played around with both power_method and power_profile and neither of them do have any effect on the clocks.

      Watching /sys/kernel/debug/dri/0/radeon_pm_info verifies this, the informations stay static all the time.
      Same here. I'm using a ATI Radeon Mobility HD2400. I tried dynpm, profile/low and profile/auto (and unplugged AC cord). No effects except random hangs.

      I can't post any debugging messages as drm.debug=1 is too verbose.

      Comment


      • #13
        I downgraded to commit c5cba545e48e4a565a7ae55e7b6affe391b36195:
        http://git.kernel.org/?p=linux/kerne...6affe391b36195

        This version works for me:
        Code:
        > cat /sys/kernel/debug/dri/0/radeon_pm_info 
        state: PM_STATE_ACTIVE
        default engine clock: 750000 kHz
        current engine clock: 250000 kHz
        default memory clock: 799000 kHz
        current memory clock: 796870 kHz
        PCIE lanes: 16
        Engine stays at 250MHz most of the time, some 500MHz runaway values - but it mostly stays in low power mode. Amount of "not in vbl for pm change" messages is minimal.

        Comment


        • #14
          Originally posted by LiquidAcid View Post
          I downgraded to commit c5cba545e48e4a565a7ae55e7b6affe391b36195:
          http://git.kernel.org/?p=linux/kerne...6affe391b36195
          My system still hangs even with the about version + dynpm. I see a lot of "not in vbl" messages.

          Comment


          • #15
            Is there any other word than awesome for this?
            Thanks to the devs!

            Comment


            • #16
              Originally posted by Adarion View Post
              Is there any other word than awesome for this?
              Thanks to the devs!
              Heh, I presume you haven't looked at the above posts stating that this doesn't work for some people Awesome would be a bit far-fetched imho

              Comment


              • #17
                Originally posted by Neuro View Post
                Awesome would be a bit far-fetched imho
                Yeah, but these are initial patches, untested on a broad variety of platforms. And if you have to use latest git code from everything ... well things tend to break there. This is a reason for me why I mainly use releases that are declared as "stable". Have to wait longer but save me the troubles on a productive system.

                Comment


                • #18
                  Will this hit next version of Ubuntu?
                  Could be awesome to have opengl 2.1 and pm out of the box.

                  Comment


                  • #19
                    I wrote simple tool for power management on radeon

                    supports both dialog(curses/terminal GUI) and Xdialog(X11/gtk GUI)

                    http://carme.pld-linux.org/~evil/rad..._scripts/radpm

                    All it needs is ksh(you can change it to bash/zsh safely), sed and dialog/Xdialog

                    Comment


                    • #20
                      Originally posted by Adarion View Post
                      Is there any other word than awesome for this?
                      Thanks to the devs!
                      the real awesomeness is that some people benchmark this stuff again fglx and the opensource driver win means lower powerconsuming than fglrx.


                      thats the REAL AWESOME ;-)

                      Comment

                      Working...
                      X