Thursday 8 April 2010

Saving power on my HPMini (revisited on Lucid)

A while ago I blogged about power saving on my HPMini. Well, today I'm revisiting this using today's Lucid daily and I'm pleased to see that I can save a small amount of power with Lucid compared to Karmic. However, the difference is small - the inaccuracies of estimating power consumption using ACPI may be more significant.

My steps were as follows:

0. Run powertop for about 15 minutes on battery power and note down the recommended power saving tricks.

1. Blacklist Bluetooth, since I don't use this and it really sucks a load of power. To do so, add the following to /etc/modprobe.d/blacklist.conf

blacklist btusb

2. Enable HDA audio powersaving:

echo 1 > /sys/module/snd_hda_intel/parameters/power_save

3. Increase dirty page writeback time:

echo 1500 > /proc/sys/vm/dirty_writeback_centisecs

4. Disable the webcam driver (not sure exactly if this saves much power), add the following to /etc/modprobe.d/blacklist.conf

blacklist uvcvideo

5. Turn down the screen brightness (this saves 0.3 Watts)

6. Disable desktop effects to save some power used by compositing.

7. Disable cursor blinking on the gnome terminal to save 2 wakeups a second, using:

gconftool-2 --type string --set /apps/gnome-terminal/profiles/Default/cursor_blink_mode off

I managed to push the power consumption down to ~6.5 Watts which is an improvement of nearly 1 Watt from my tests on Karmic. Not bad, but I don't fully trust the data from the battery/ACPI and I really need to see how much extra battery life these tweaks give me when I'm using the machine on my travels.

On an idle machine I'm seeing > ~99.0% C4 residency, which is quite acceptable. Running powertop is always leads to insights to saving power, so kudos to Intel for this wonderful utility.

2 comments:

  1. Rather than blacklisting the bluetooth driver, I just click on the panel icon and "Turn off bluetooth" which should power down the radio.

    Does blacklisting the module save more power for you?

    ReplyDelete
  2. @mdz, I went for the sledgehammer to crack a nut approach.

    ReplyDelete