Find the answer to your Linux question:
Page 2 of 2 FirstFirst 1 2
Results 11 to 20 of 20
I was thinking that aggressive throttling may effectively lower the cpu core temperature when idle. Had the same thought and it proved accurate. Before installing cpufreq, CPU idled at 3GHz ...
Enjoy an ad free experience by logging in. Not a member yet? Register.
  1. #11
    Linux Newbie
    Join Date
    Sep 2007
    Posts
    218

    I was thinking that aggressive throttling may effectively lower the cpu core temperature when idle.
    Had the same thought and it proved accurate. Before installing cpufreq, CPU idled at 3GHz and temp was 37C. Set to On Demand, CPU now idles at 800MHz and is 32C.

    For CPU capabilities:
    Code:
     cpufreq-info
    (this is a Debian machine)
    For Debian:
    Code:
    cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
    and
    cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
    See /usr/share/doc/cpufreq and cpufrequtils for info

  2. #12
    Linux Newbie
    Join Date
    Dec 2010
    Posts
    146
    Can you do normal frequency scaling (800, 1600 mhz etc...)

  3. #13
    Linux Newbie
    Join Date
    Sep 2007
    Posts
    218
    Can you do normal frequency scaling (800, 1600 mhz etc...)
    Not sure what 'normal' frequency scaling is (there are five governors/settings to choose from) and I don't actually do anything. My CPU's available frequencies are 3.00GHz, 2.30GHz, 1.80GHz and 800MHz. I chose OnDemand; the setting defaults to the lowest frequency (800MHz) and scales up as high as needed, whenever its needed. For example, open a simple GUI text editor like leafpad and one core might scale to 1.80GHz; open a VM in VirtualBox and two or three cores might scale to 3.00GHz for a while.

  4. #14
    Linux Newbie
    Join Date
    Dec 2010
    Posts
    146
    Then I think you don't have T state support for your CPU.

    There're other ways you can save power like with hdparm, and maybe disabling caching will help.

  5. #15
    Linux Newbie
    Join Date
    Sep 2007
    Posts
    218
    Then I think you don't have T state support for your CPU.
    Had to look that up and I'm still not sure what T state support is... is it auto-throttling a CPU when its temp goes above a mfg-set limit? Also, what leads you to think this CPU doesn't have T state support?

    There're other ways you can save power like with hdparm
    Thought about doing that to spin down the 2nd HD when not used; its for multimedia stuff. With almost daily use, though, would it save much power?

    maybe disabling caching will help
    How much power would that actually save?

  6. #16
    Linux Newbie
    Join Date
    Dec 2010
    Posts
    146
    Quote Originally Posted by fanderal View Post
    Had to look that up and I'm still not sure what T state support is... is it auto-throttling a CPU when its temp goes above a mfg-set limit? Also, what leads you to think this CPU doesn't have T state support?


    Thought about doing that to spin down the 2nd HD when not used; its for multimedia stuff. With almost daily use, though, would it save much power?


    How much power would that actually save?
    The cpu will only work x% of the time, i.e. it'll turn off in between to save power. So this's different from scaling down the core frequency, my bets are that under low loads, this will save power.

    Laptop HDDs I think consume 5 to 7 W.

    Caching -- I'm not sure but I think if less parts of the physical ram is occupied, only the occupied part of it has to be refreshed. By default Linux caches a lot to physical memory.

  7. #17
    Linux Newbie
    Join Date
    Sep 2007
    Posts
    218
    Your comments got me curious, partly because its an area I'm not familiar with, and partly because I wondered how you could know what my CPU doesn't support without knowing the CPU's make and model. Anyway, I googled s'more and have a little better understanding.

    The cpu will only work x% of the time, i.e. it'll turn off in between to save power. So this's different from scaling down the core frequency, my bets are that under low loads, this will save power.
    I agree, its not the same as frequency scaling. What you've described is skipping clock cycles, which doesn't reduce frequency or voltage, just throughput. Its when a 'no operation' signal (google "CPU nop") is passed to the CPU x times out of every y cycles, telling the CPU to idle (not turn off). It doesn't save power; the T state's purpose is to save the CPU from burning out. I found definitions for T and other states here: LessWatts.org - Saving Power on Intel systems with Linux

    This isn't a laptop. Its a desktop and the CPU's frequency and the corresponding power consumption do actually change. BTW, if the 'no operation' signals the CPU to idle, what would account for a 5C drop in the CPU's temp while idling? My reply to the OP was about lowering the CPU temp, which was my purpose, too, and I don't think that can happen without true frequency scaling.

    Laptop HDDs I think consume 5 to 7 W.
    The question was how much power might be saved by using hdparm to spin down a hard disk... in this instance, spinning a desktop HD down maybe two days a week -vs- not spinning it down at all. I've no idea what it would save but I doubt its much.

    Caching -- I'm not sure but I think if less parts of the physical ram is occupied, only the occupied part of it has to be refreshed.
    Well, I guess that makes two of us 'cause I dunno how much power might be saved.

  8. #18
    Linux Newbie
    Join Date
    Dec 2010
    Posts
    146
    In this source --

    LessWatts.org - Saving Power on Intel systems with Linux

    They did not state if T states will save power if the CPU is idling. I'll check this out.

  9. #19
    Linux Newbie
    Join Date
    Dec 2010
    Posts
    146
    Well, I did some practicals with T states, it seems like it increased the temperature and so power consumption, even when I'm idle.

    For desktop systems -- why not get more fans?

  10. #20
    Linux Newbie
    Join Date
    Sep 2007
    Posts
    218
    They did not state if T states will save power if the CPU is idling.
    This seems pretty clear (from LessWatts):
    The purpose of the T state is to prevent the CPU from overheating by lower its temperature.

    T-states forcefully introduce idle cycles in the processor
    Idle cycles do not get into power saving C-states
    Well, I did some practicals with T states, it seems like it increased the temperature and so power consumption, even when I'm idle.
    If by "practicals" you mean you tested only your CPU's T state execution, I doubt you'd have seen those results. When searching, I ran across a video (didn't save the url) where it was inadvertantly demonstrated. It showed a monitor with a full screen, CPU-intensive video game and a small window showing CPU temps. The game was smooth and responsive until the temp climbed to something near 90C. At that point, two things happened... the temps began to level off and then slightly decrease, and the game became less smooth and responsive.

    Why? I think the T state introduced idling into the CPU cycles, and with reduced stress on the CPU, it began to cool. It also degraded the game's video because the CPU was no longer functioning at full capacity... idle cycles don't process anything so the CPU was not able to process the video quick enough to run smoothly.

    FWIW, I wouldn't recommend this kinda testing. If it works, it'll put a fair amount of stress on the CPU. If for some reason it doesn't work and the T state doesn't kick in, it means buying a new CPU. Not very good odds

    I think we've taken this thread far enough off topic. Maybe start your own thread to discuss power saving techniques/methods? Others' comments would certainly be helpful.

Page 2 of 2 FirstFirst 1 2

Posting Permissions

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