Re: Power consumption HZ100, HZ250, HZ1000: new numbers

From: James Bruce (bruce_at_andrew.cmu.edu)
Date: 08/03/05

  • Next message: Kyle Moffett: "Re: Calling suspend() in halt/restart/shutdown -> not a good idea" 
    Date:	Wed, 03 Aug 2005 12:47:56 -0400
To: ambx1@neo.rr.com

    (Sorry all, but after receiving about 5 similar messages I'm going to
    make one last reply.)

    ambx1@neo.rr.com wrote:
    > Also, my understanding was that when we properly support usb suspend,
    > this won't be an issue anyway for much usb hardware. I think it's
    > possible to put some mice to sleep when there isn't any motion and
    > then wakeup later.

    By the time we properly support USB suspend, it won't matter because the
    dynamic tick patch will likely have been integrated. An argument for
    why we should change a value in the future is pointless when the
    question is what the value should be right now.

    > 4.4% savings may not be much, but these things do add up.

    A 300% increase in minimum sleep latency adds up quite quickly.

    The point that people joining this thread keep missing is that we're
    making a change that:
       (a) offers a small benefit to laptop users
       (b) messes up other uses such as video
       (c) is likely to be completely obsolete by 2.6.14

    > For a laptop's workload, I think this is worth it.

    Good, so on your laptop go choose 100Hz. You already have to configure
    your kernel to change values from their defaults anyway, otherwise you
    won't see any of that 4.4% savings. However, please don't screw up
    video and interactivity (1) for everyone who uses default values on
    their desktops in order to get your savings.

    (1) http://article.gmane.org/gmane.linux.kernel/319124/

    Rather like the famous gcc 2.96, this is something that will confuse
    users for some time to come. Just about every video app will need an
    FAQ entry to say why 2.6.13 doesn't work as well and drops frames while
    2.6.(x!=13) works just fine. Unless of course distros read this thread
    and decide not to pick up this change; That I guess is the only reason
    I'm still posting.

    Now as Lee said, please let this thread die. We need to go work on
    dyntick and test it on as much hardware as possible and try to uncover
    any lurking bugs. If you care about saving power or multimedia, you
    should test it too (think ~10% savings rather than 4.4%).

      - Jim Bruce
    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

  • Next message: Kyle Moffett: "Re: Calling suspend() in halt/restart/shutdown -> not a good idea"