Results 1 to 7 of 7

Thread: IBM: Memory Power Management Savings Measurable

  1. #1
    Join Date
    Jan 2007
    Posts
    14,646

    Default IBM: Memory Power Management Savings Measurable

    Phoronix: IBM: Memory Power Management Savings Measurable

    An IBM Linux Technology Center has shared that enabling experimental memory power management within the Linux kernel has dropped one of their test system's power consumption by about 2.6% but it's likely even more with experimental hardware...

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

  2. #2
    Join Date
    Sep 2009
    Posts
    318

    Default Seems like a waste of energy!

    I don't see a big draw here. Sounds like a lot of talent going into a feature that only supports a few specialized machines.

  3. #3
    Join Date
    Jan 2009
    Location
    Outthere, NSW, Australia
    Posts
    350

    Default

    Quote Originally Posted by wizard69 View Post
    I don't see a big draw here. Sounds like a lot of talent going into a feature that only supports a few specialized machines.
    Heh. Big draw and power savings. You jest!

  4. #4
    Join Date
    Apr 2008
    Location
    Zagreb, Croatia
    Posts
    102

    Default

    Quote Originally Posted by wizard69 View Post
    I don't see a big draw here. Sounds like a lot of talent going into a feature that only supports a few specialized machines.
    Well, in a 10MW datacenter, a 1% power saving over 10 years
    translates to couple of millions of USD. It may not sound much,
    but electricity bill is the biggest chunk of TCO in a data center.

  5. #5
    Join Date
    Feb 2008
    Location
    Linuxland
    Posts
    5,070

    Default

    "few specialized machines"? This feature is already supported on some ARM boards and is coming to x86.

  6. #6
    Join Date
    Oct 2006
    Location
    Israel
    Posts
    585

    Default

    Quote Originally Posted by wizard69 View Post
    I don't see a big draw here. Sounds like a lot of talent going into a feature that only supports a few specialized machines.
    A normal 4 socket Intel Xeon machine and/or AMD Opteron machine (a 20,000$) server usually have ~40+ cores and anywhere between 128GB and 1TB (!) of memory, and I'm not talking about a 100+K$ mainframe machine like the IBM Power series.
    Plus, keep in mind that as you increase the amount of memory (I've got a number of machines with 512GB RAM), the power saving should increase (at least in theory).

    - Gilboa
    DEV: Intel S2600C0, 2xE52658V2, 32GB, 4x2TB, GTX780, F20/x86_64, Dell U2711.
    SRV: Intel S5520SC, 2xX5680, 36GB, 4x2TB, GTX550, F20/x86_64, Dell U2412..
    BACK: Tyan Tempest i5400XT, 2xE5335, 8GB, 3x1.5TB, 9800GTX, F20/x86-64.
    LAP: ASUS N56VJ, i7-3630QM, 16GB, 1TB, 635M, F20/x86_64.

  7. #7
    Join Date
    Jan 2009
    Location
    Outthere, NSW, Australia
    Posts
    350

    Default

    All I can say is 2% here, 5% there, adds up real quick in the sceme of things. And he's done it simply ny improving managing the hardware via software. Imagine if the rest of tje dev team made single digit percentile improvements, the cost of operation would be massively reduced. Reduced power co sumption means reduced cooling and failing , reducing costs alround. Its not rock'n'roll science, but itll do its part for tje planet and the pocket!

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •