Announcement

Collapse
No announcement yet.

Talk Of A "Massive Power Regression" In Linux 3.5

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

  • Talk Of A "Massive Power Regression" In Linux 3.5

    Phoronix: Talk Of A "Massive Power Regression" In Linux 3.5

    For at least some hardware, it looks like the Linux 3.5 kernel has regressed and is burning through noticeably more power than its predecessor...

    http://www.phoronix.com/vr.php?view=MTE0OTg

  • #2
    I can see these power issues are becoming a major problem for Linux in general. It's hard to say to Windows users: "hey, this software is free, but you'll get an hour less of battery life". This is simply not acceptable, no matter how you twist it.

    I know the troubles involved in making progress (uncooperative companies, undocumented protocols, closed-off hardware, etc.), but this is something that we as a community need to put pressure on. I don't know who, or which influential company, will need to fire the first salvo, but it's something that's got to be done to make hardware manufacturers realize the seriousness of the issue.

    Comment


    • #3
      Ah. I was wondering when we would get another power regression. Anyone want to make bets on how many kernel versions will pass before it is fixed? My bet is 4.

      Comment


      • #4
        Originally posted by ua=42 View Post
        Ah. I was wondering when we would get another power regression. Anyone want to make bets on how many kernel versions will pass before it is fixed? My bet is 4.
        i bet some lazy kernel dev will push the big red button named "fix da driver" in the next month TROLL

        is impressive like some ppl like minimize or have fun [blessful ignorance] of how massive can be fix some "trivial" bug in an inmensively complex piece of software like a kernel driver and the affected subsystems

        Comment


        • #5
          Switching between 3.2 and 3.4 (debian kernels) I have the impression that already 3.4 chews through more power than 3.2. Power efficiency really needs more attention from guys like Intel or RedHat or google.

          Comment


          • #6
            TBH i was expecting that someone (someone on the kernel team that is) would be testing for that kind of regressions over the rc cycle.

            Comment


            • #7
              Originally posted by not.sure View Post
              Switching between 3.2 and 3.4 (debian kernels) I have the impression that already 3.4 chews through more power than 3.2. Power efficiency really needs more attention from guys like Intel or RedHat or google.
              well you can contribute too generating a bug in bugzilla so the kernel developers can try to find out where is the issue[note sometimes is distros that screw some pm settings].

              try to describe fully wich distro/kernel/hardware you use when posting

              Comment


              • #8
                Originally posted by 89c51 View Post
                TBH i was expecting that someone (someone on the kernel team that is) would be testing for that kind of regressions over the rc cycle.
                well imagine is happens only with atom cpu but with a bugged motherboard from zotac to put an example but it doesn't happen with an intel board, so you need that ppl with this specific setup report the bug [not just whine about it in a forum unrelated to kernel development], unlike many ppl here think kernel developer don't practice witchcraft or divination and you can't realisticly expect them to test every possible combination of hardware know to men without have release an stable kernel every 15 years

                so yes reporting those bugs in bugzilla[google: kernel bugzilla] make a huge diference and is actually the bigger step needed to fix those issues

                Comment


                • #9
                  Originally posted by jrch2k8 View Post
                  well imagine is happens only with atom cpu but with a bugged motherboard from zotac to put an example but it doesn't happen with an intel board, so you need that ppl with this specific setup report the bug [not just whine about it in a forum unrelated to kernel development], unlike many ppl here think kernel developer don't practice witchcraft or divination and you can't realisticly expect them to test every possible combination of hardware know to men without have release an stable kernel every 15 years

                  so yes reporting those bugs in bugzilla[google: kernel bugzilla] make a huge diference and is actually the bigger step needed to fix those issues
                  Well that's totally the wrong way of doing things: release broken, non tested stuff and try to fix them sometime in the future.

                  Windows does not have this issue because they do things the other way: test test test, before releasing and is why they release a preview release and a bunch of other releases prior to final.

                  example:
                  http://en.wikipedia.org/wiki/Windows_8
                  -Several Milestones
                  -Developer Preview
                  -Consumer Preview
                  -Release Preview
                  -Release to Manufacturing
                  -and then General availability (if ready)

                  This "we need to release" a kernel (and/or a distro) every 6 months (because we want to) with limited to no testing.

                  These fixed dates don't make much sense, and is why we always have these regressions. And then they say "don't worry it will probably be fixed on next releases" and then a release after they break it again and the cycle of breakage goes on forever.

                  If windows had the same "we don't care" linux release model it would be a broken mess that fails every 6 months.

                  In conclusion: please Release if ready !

                  "Insanity: doing the same thing over and over again and expecting different results."
                  -- Albert Einstein
                  Last edited by madjr; 07-30-2012, 06:07 PM.

                  Comment


                  • #10
                    Originally posted by madjr View Post
                    In conclusion: please Release if ready !
                    Linux is used quite literally everywhere. You can't delay the entire Linux release because of some bug that just affects some specific case. It's up to distribution to choose which kernel releases to use and they can even patch these problems themselves before hand. You can't compare Linux to Windows because Linux is simply a kernel. New major version of Linux is released once in 60-70 days btw.

                    Comment


                    • #11
                      Originally posted by madjr View Post
                      Well that's totally the wrong way of doing things: release broken, non tested stuff and try to fix them sometime in the future.

                      Windows does not have this issue because they do things the other way: test test test, before releasing and is why they release a preview release and a bunch of other releases prior to final.

                      example:
                      http://en.wikipedia.org/wiki/Windows_8
                      -Several Milestones
                      -Developer Preview
                      -Consumer Preview
                      -Release Preview
                      -Release to Manufacturing
                      -and then General availability (if ready)

                      This "we need to release" a kernel (and/or a distro) every 6 months (because we want to) with limited to no testing.

                      These fixed dates don't make much sense, and is why we always have these regressions. And then they say "don't worry it will probably be fixed on next releases" and then a release after they break it again and the cycle of breakage goes on forever.

                      If windows had the same "we don't care" linux release model it would be a broken mess that fails time after time.

                      In conclusion: please Release if ready !

                      "Insanity: doing the same thing over and over again and expecting different results."
                      -- Albert Einstein
                      this exist in the linux kernel too but the biggest difference is that microsoft[even more mac] only test on a selected set of hardware the rest of the drivers are tested by the hardware vendors and this problem is not unique to linux either all windows version has suffered from very nasty regression and bugs too[dunno in which world you live in tho ] hence that is why you have drivers hotfixes and service packs/critical updates all the time[they don't put it there for fun you know]

                      btw linux have LTS releases wich are very well tested[3.2 series] and is used in all commercial / stable distros[rhel/debian/ubuntu LTS/etc], so if you want a stable/tested software stick with a stable distro but if you wanna live bleeding edge don't whine and go report bugs if you find them

                      note that kernel releases are not releases or final releases like in other OSes but more like bleeding edge but usable enough for test the next stuff, distros are the ones that test those kernel to death and stick to them for a nice time until the next one is ready for produccion systems[named red hat and debian].

                      so you can't compare linux kernel to windows[one is a kernel and the other is an OS], the correct would be redhat/debian to windows[both are OSes one with a linux kernel the other with an NT kernel] and in that sense redhat and debian are very stable [more than windows in many cases] and enterprise proven

                      for your information windows do the same with the NT kernel you just can't get the next release since is closed to microsoft employees only while linux dev happens realtime and openly

                      Comment


                      • #12
                        Originally posted by not.sure View Post
                        Switching between 3.2 and 3.4 (debian kernels) I have the impression that already 3.4 chews through more power than 3.2. Power efficiency really needs more attention from guys like Intel or RedHat or google.
                        Why don't the debian devs try and fix it?
                        Red Hat and Google and Intel have their own priorities and power management might not be at the top of stack. Debian, being composed of volunteers, can more easily choose what they work on.

                        Comment


                        • #13
                          Originally posted by jrch2k8 View Post
                          this exist in the linux kernel too but the biggest difference is that microsoft[even more mac] only test on a selected set of hardware the rest of the drivers are tested by the hardware vendors and this problem is not unique to linux either all windows version has suffered from very nasty regression and bugs too[dunno in which world you live in tho ] hence that is why you have drivers hotfixes and service packs/critical updates all the time[they don't put it there for fun you know]

                          btw linux have LTS releases wich are very well tested[3.2 series] and is used in all commercial / stable distros[rhel/debian/ubuntu LTS/etc], so if you want a stable/tested software stick with a stable distro but if you wanna live bleeding edge don't whine and go report bugs if you find them

                          note that kernel releases are not releases or final releases like in other OSes but more like bleeding edge but usable enough for test the next stuff, distros are the ones that test those kernel to death and stick to them for a nice time until the next one is ready for produccion systems[named red hat and debian].

                          so you can't compare linux kernel to windows[one is a kernel and the other is an OS], the correct would be redhat/debian to windows[both are OSes one with a linux kernel the other with an NT kernel] and in that sense redhat and debian are very stable [more than windows in many cases] and enterprise proven

                          for your information windows do the same with the NT kernel you just can't get the next release since is closed to microsoft employees only while linux dev happens realtime and openly
                          Ok , so it's alright for the kernel devs to barely QA and it basically should be the distros job ? .....

                          Am getting the impression that there is a good lack of communication and coordination. Everyone wants to wipe their hands clean and point fingers to the others. Seems most open source projects have this attitude lately (including gnome). Everyone just wants to do their own thing.

                          I call this Open code, Closed minds.
                          Last edited by madjr; 07-30-2012, 11:19 PM.

                          Comment


                          • #14
                            Originally posted by madjr View Post
                            Ok , so it's alright for the kernel devs to barely QA and it basically should be the distros job ? .....
                            Yes.

                            You are not employing the kernel developers.
                            You are not giving free hardware samples to the kernel developers.
                            You are not even willing to provide bug reports to the kernel developers.

                            Also, no one is forcing you to use Linux, Gnome or whatever it is you don't like.

                            Comment


                            • #15
                              If it doesn't work for you, just don't bloody use it and shut up, or file a bug report and shut up.

                              Comment

                              Working...
                              X