Announcement

Collapse
No announcement yet.

More Radeon Power Management Improvements

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

  • Yes, evergreen is at the same level as previous generations.

    Comment


    • Can we somehow change the pm settings as user? Or would it be "secure" enough letting a plasma applet change the values as root?

      I am considering programming a plasma widget to change the pm settings.

      Comment


      • Originally posted by HokTar View Post

        Check your current state:
        cat /sys/kernel/debug/dri/0/radeon_pm_info

        Now you can recheck if your power state changed to the desired one:
        cat /sys/kernel/debug/dri/0/radeon_pm_info
        Thanks for the mini guide, but my kernel doesn't seem to have any directories within the debug directory. Well it seems that I can set the power mode anyway.

        Using latest stable 2.6.35.

        Comment


        • Originally posted by tball View Post
          Thanks for the mini guide, but my kernel doesn't seem to have any directories within the debug directory. Well it seems that I can set the power mode anyway.

          Using latest stable 2.6.35.

          You need to mount debugfs somewhere, like this :
          Code:
          mkdir /debugfs
          mount -t debugfs debugfs /debugfs
          cat /debugfs/dri/0/radeon_pm_info

          Well, for me it works...(r7xx, radeon 4550)...but it seems to slow to get back to normal temperature...I mean, even in low mode, the fan need at least 5 minutes to spin down again after playing a video.
          And why are the PCIe lanes are always equals to 16 whatever the profile mode...?

          Comment


          • The pcie lanes are only adjusted if the power profile requests it; not all profiles do.

            Comment


            • And which profile does ? I thought "low" would be the one...
              Just a question : is development for powersaving features done ? I mean, it still needs improvement, doesn't it ? Still not as good as the catalyst powerplay....

              Comment


              • Originally posted by Gaius View Post
                And which profile does ? I thought "low" would be the one...
                Just a question : is development for powersaving features done ? I mean, it still needs improvement, doesn't it ? Still not as good as the catalyst powerplay....
                I believe the profiles just look up settings from the cards BIOS, which can have card/vendor specific settings in it. So if your cards low-power profile is set to X Mhz core, Y Mhz mem, and 16 lanes, then that's what the driver goes to.

                Comment


                • That's a very bad news indeed...so there is no way it would have someday the same behavior as the catalyst driver (without reflashing the bios) ? I mean what if the only way to reduce even more the energy consumption was to reduce the pcie lanes...it must be done somehow by the catalyst driver...or the dynamic clocking algorithm is different...

                  Comment


                  • How is that bad news ? All it means is that the current open source power management doesn't do as much as fglrx power management, but that is hardly a revelation, and it says nothing about what can/will happen in the future.

                    Comment


                    • Originally posted by Gaius View Post
                      I mean what if the only way to reduce even more the energy consumption was to reduce the pcie lanes...it must be done somehow by the catalyst driver
                      Well, the code is open source, so you can reduce the number of pcie lanes yourself to find out. I played around with that a bit when power management was only in the user space driver and I did not find that reducing the number of pcie lanes resulted in lower power consumption, only to instability

                      FWIW I'm currently using fglrx at the moment and it's using all of the 16 pcie lanes with a HD 5750, regardless of the performance level.

                      Comment


                      • Originally posted by monraaf View Post
                        Well, the code is open source, so you can reduce the number of pcie lanes yourself to find out. I played around with that a bit when power management was only in the user space driver and I did not find that reducing the number of pcie lanes resulted in lower power consumption, only to instability

                        FWIW I'm currently using fglrx at the moment and it's using all of the 16 pcie lanes with a HD 5750, regardless of the performance level.
                        I already did have a look at the code in radeon_pm.c but it is a bit too complex for me
                        So you mean, the PCIe lanes are not so important..that's interesting...maybe it is the algorithm used in the radeon driver that needs improvement

                        Comment


                        • Originally posted by Gaius View Post
                          ...maybe it is the algorithm used in the radeon driver that needs improvement
                          What would you suggest, remembering that the PM code in the open source stack is perhaps 1/20th the size and complexity of the corresponding code in fglrx ?

                          Comment


                          • Originally posted by bridgman View Post
                            What would you suggest, remembering that the PM code in the open source stack is perhaps 1/20th the size and complexity of the corresponding code in fglrx ?
                            less code means higher code quality.
                            and less security holes.
                            and higher stability.

                            i think the FGLRX only have more code as the linux kernel because the linux kernel delete bad and obsoled code and fglrx collects all code.

                            more core is just a 'number'

                            Comment


                            • Originally posted by bridgman View Post
                              What would you suggest, remembering that the PM code in the open source stack is perhaps 1/20th the size and complexity of the corresponding code in fglrx ?
                              I guess it's mostly about determining the absolute minimum settings on each card. Many laptop users would probably be perfectly happy with that? That's a rather tricky thing though if the card BIOS does not contain the absolute minimum settings. Should the driver still rely on them or ship with overrides users could manually enable?

                              Comment


                              • Originally posted by Gaius View Post
                                maybe it is the algorithm used in the radeon driver that needs improvement
                                In all honesty I don't really have any complaints about the power management in the open source radeon driver. It works like a charm for me with a HD 5750, so I don't see what exactly needs improvement.

                                System power consumption Idle Desktop:
                                -----------------------------------------
                                fglrx : 62 W
                                radeon: 65 W

                                GPU Temperatures Idle Desktop:
                                -----------------------------------------
                                fglrx: 40 C
                                radeon: 42/43 C


                                The only thing that I'm missing is something to manually control the fan, so I can make the thing just a little more quiet.

                                Comment

                                Working...
                                X