Recent Linux kernels waste energy in certain situations
The Phoronix web site, which focuses on the interaction between Linux and hardware components, has reportedly found in tests that some notebooks running current Linux kernels require 10 to 30 per cent more power than those running older versions of Linux. The web site has so far published five articles on the subject, some of them several pages long, and provides links to these articles in the first section of the latest part.
In its tests, Phoronix was reportedly able to identify two different situations where the Thinkpad notebooks that were used for testing consumed noticeably more power with recent kernel versions. One of the problems reportedly only exists from 2.6.38; the other was found with one of the Thinkpads from kernel version 2.6.35.
An Ubuntu developer has also observed increased power consumption, and therefore a shorter battery life, in preview versions of Ubuntu 11.04, and has created an Ubuntu bug report about the problem. In this thread, other testers have reported similar experiences; the main factor for the increased power consumption often appears to be the processor, which is prevented from remaining in the power-saving short-term sleep modes by an increased load. According to the bug report, which has been given a "high" importance rating and is being followed by about 80 users, the Ubuntu developers hope to fix the problem with an update.
The causes for the increased power consumption have not been identified; it is also unclear whether all the testers are looking at the same problem and which types of systems are affected; often, such problems only exist with certain processors, graphics cards or hardware combinations. Phoronix has, for several days, been looking to identify the kernel changes that were made when the increased power consumption started to register with its test devices. Only once these changes have been found will it be possible to judge whether the problems affect a few, many or even all Linux users. We haven't been able to find any discussions about this problem on the usual kernel developer mailing lists.
It is not the first time that Phoronix has found a similar problem and caused a lot of commotion. Several kernel development veterans in key positions have previously made critical comments about the web site (1, 2, 3, 4) in this context. Both The H's associates at c't and heise open have also found similar, potentially serious Linux problems in the past. On closer inspection, however, good reasons that may not have been very obvious did exist for many of these problems. Many other issues were quickly resolved by the distribution or kernel developers once their causes had been identified; the current case will probably be no exception.