The Linux Kernel Power Problems On Older Desktop Hardware

Written by Michael Larabel in Software on 22 June 2011 at 11:09 AM EDT. Page 2 of 7. 53 Comments.

The first result is using the OpenBenchmarking.org battery-power-usage test profile that's measuring the Wattage while the system is idling at the desktop, while still idling but with the display disabled via DPMS, and then the display re-enabled and a brief 1080p video playback is run via MPlayer with the X-Video interface.

These results are quite interesting and it shows the pre-2.6.34 kernels are consuming a vicious amount of power. The Linux 2.6.32 and 2.6.33 kernels while idling were consuming as much power as the 2.6.34 and later kernels while under full load. The Linux 2.6.34 kernel marks a big improvement in power management for this old Pentium 4 + Radeon 9200 PRO system. In terms of the power regressions that we have been tracking down, they are confirmed by this test. The well-known Linux 2.6.38 power regression is hitting this system along with the other big power regression that was introduced in the Linux 2.6.35 kernel.

While the Linux 2.6.34 kernel improves power management a lot over the 2.6.32/2.6.33 releases, with the Linux 2.6.35 kernel the average power consumption is 59 Watts compared to 54 Watts in the Linux 2.6.34 kernel. Then with the 2.6.38 regression, the average power consumption during this test is 60 Watts on the 2.6.37 release and up to 65 Watts on average with the affected 2.6.38 kernel. The average power consumption with the latest Linux 3.0 development kernel is up to 70 Watts.


Related Articles