PowerManagementASPM

Differences between revisions 1 and 7 (spanning 6 versions)
Revision 1 as of 2011-11-12 12:28:37
Size: 1404
Editor: colin-king
Comment:
Revision 7 as of 2011-11-12 18:49:18
Size: 2337
Editor: hugwijst
Comment:
Deletions are marked like this. Additions are marked like this.
Line 9: Line 9:
[[http://zinc.canonical.com/~cking/powermanagement/mjgaspmfix/oneiric|Oneiric Kernels]]
[[http://zinc.canonical.com/~cking/powermanagement/mjgaspmfix/precise|Precise Kernels]]
 * [[http://zinc.canonical.com/~cking/powermanagement/mjgaspmfix/oneiric|Ubuntu Oneiric 11.10 Kernels]]
 * [[http://zinc.canonical.com/~cking/powermanagement/mjgaspmfix/precise|Ubuntu Precise 12.04 Kernels]]

For power measurement, we currently recommend using the older version of powertop 1.13 as this clearly displays the ACPI estimated power consumption. A .deb for this can be found [[https://launchpad.net/~colin-king/+archive/powermanagement|here]]. We recommend running powertop as follows:

 * Boot the machine, start powertop 1.13
 * Wait until you get an ACPI estimated power consumption value.
 * Gather 15 measurements (it is updated regularly).
 * Discard the 5 results.
 * Take average of the last 10 results.
Line 16: Line 24:
 * Run powertop for 10 minutes on an idle system and note power consumption  * Run powertop on an idle system and note power consumption
Line 21: Line 29:
 * Run powertop for 10 minutes on an idle system and note power consumption  * Run powertop on an idle system and note power consumption
Line 26: Line 34:
||Example||68ET27WW||64||3.1.0-2-generic||23W||3.1.0-2-generic||19W|| ||Lenovo 3000N200||68ET27WW||64||3.1.0-2-generic||24.3W||3.1.0-2-generic||25.4W||
||HP Mini 210||F0.2||64||3.0.0-12-generic||14.3W||3.0.0-13-generic||8.7W||
||HP ProBook 5420s||68AZZ Ver. F.0F||64||3.0.0-12-generic||25.06W||3.0.0-13-generic||25.66W||

Notes: Power estimation using ACPI battery data is not strictly accurate. We may see variation in data. It may be worthwhile repeating the test several times so we can see statistical variations in the gathered data.

ASPM PCIe Bug

As noted by Phoronix ASPM PCIe power saving has been problematic and a recent patch by Matthew Garrett has re-worked this to clear the ASPM state on all devices on a successful handoff of PCIe control to the OS.

We therefore are providing patched kernels carrying Matthew's fix for testing purposes only to see how well this works across a broad range of machines. Kernels can be found in:

For power measurement, we currently recommend using the older version of powertop 1.13 as this clearly displays the ACPI estimated power consumption. A .deb for this can be found here. We recommend running powertop as follows:

  • Boot the machine, start powertop 1.13
  • Wait until you get an ACPI estimated power consumption value.
  • Gather 15 measurements (it is updated regularly).
  • Discard the 5 results.
  • Take average of the last 10 results.

Testing procedure:

  • Ensure laptop is fully charged
  • Unplug AC power
  • Run powertop on an idle system and note power consumption
  • Plug in AC power
  • Install the new kernel containing the ASPM fix
  • Ensure laptop is fully charged
  • Unplug AC power
  • Run powertop on an idle system and note power consumption
  • Gather the BIOS version using: sudo dmidecode -t 0 | grep "Version:"
  • Add results to the table below:

Machine

BIOS

32/64 bit kernel

Original Kernel

Original Kernel Power

ASPM fix Kernel

ASPM fix Power

Lenovo 3000N200

68ET27WW

64

3.1.0-2-generic

24.3W

3.1.0-2-generic

25.4W

HP Mini 210

F0.2

64

3.0.0-12-generic

14.3W

3.0.0-13-generic

8.7W

HP ProBook 5420s

68AZZ Ver. F.0F

64

3.0.0-12-generic

25.06W

3.0.0-13-generic

25.66W

Notes: Power estimation using ACPI battery data is not strictly accurate. We may see variation in data. It may be worthwhile repeating the test several times so we can see statistical variations in the gathered data.

Kernel/PowerManagementASPM (last edited 2012-02-19 14:59:07 by costing)