Share: |

I have recently been talking (at least tangentially) about intuition and computers, most notably in “Isolation Experimentation” and “Spaced Out”. As computer professionals we often base our thinking on how computers are and how they should work off an unconscious synthesis of the facts, and unfortunately, the opinions that we have. Often we even base prejudgment about the computer systems we work with on things that used to be true, but are not anymore. Sometimes it is guilt by association even: See ReiserFS for details.


Here was one of my intuitions: Linux is more power efficient than MS Windows.


As the Green IT person, I got to thinking about and questioning assumptions, and that one floated around in the back of my head. I pulled if forward, and looked at it, and tried to decide why I thought that (or maybe “felt that”). Clues started to assemble as I pondered it.


Mostly the idea came from laptops I have built over the years. Many times rather than deleting the MS Windows (98, NT, XP, Vista, or Win7) that was installed on it, I would make it dual boot. All kinds of laptops, from IBM, Dell, Toshiba, and Acer. Small portables like the Dell D420 or Acer Aspire One, Medium size systems like the IBM T41 and T43. Big systems like the Dell M4500 or Acer 56xx. Processors from AMD and Intel. Six cell batteries and nine cell batteries.


You name it. Whatever the laptop was, it always seemed to run longer when on battery and booted to Linux than when on battery and booted to whatever the MS Windows de Jour was. Seemed to. Very subjective.


Another utterly subjective, almost unmeasurable thing: the desktop always felt faster under Linux, which my mind mapped to it was doing less when I was not asking it to do things. It waited more quietly. I always run with the system monitor in the task bar so I can watch what the system is doing, and most of the time it is not doing much of anything.


The fans ran less when on Linux. That in and of itself would map to better battery life, since a fan uses a lot of power. A fan running is a symptom of a hotter CPU though, and more heat equals higher power drain as well. I can observe this every day now on the Dell M4500. When using Linux (Mint 10 at the moment), the fan runs less often, and far slower. When under Win7 there is a warm breeze always emanating from the left side of the system. In fact, if the fan starts running when on Linux, I start looking around to see what is doing it: Almost always some web page with a ton of active content.


I use the same web browsers, although the email is Outlook versus Evolution. I use LibreOffice / Open Office both places, though I do occasionally fire up Excel for certain spreadsheets that don't work as well yet in OO/LO. I tend to have about the same number of applications up at the same time. That stands to reason as I use Linux as my desktop, and I use it to do the same things I would do when Win7 is up.


Still: Subjective. Indicative even. Not evidence. Not controlled study. Bias is noted and admitted.


As both the Green IT person and the "Adventures in Linux" writer, someone sent me a link to this:


This is data showing Linux does in fact use less power than MS Windows on servers, and they detail their testing methodologies.  You can argue that the power profiles chosen favored MS Windows, since they did not use the most power saving settings, or that they favored Linux because they did not user powercfg.exe on Windows to try and squeeze everything down to to the lowest possible settings. Etc. Etc. Certainly the conversation in the comments section showed much disagreement about the methods chosen.


At the end of it all, Linux, specifically RedHat 5.1, was 12% more power efficient overall.


What stood out for me was that (even though this was 1.6 years ago) they were testing ancient versions of Linux (RHEL 5.1?) On ancient hardware (Dell 1950's? Really?). This was all pre the Linux tickless kernel! RHEL6's major power saving features were not in play at all! The huge advancements in recent years on the hardware side, where entire servers can be quiesced were not in scope. On and on.


To balance it a bit, this test was Windows 2008, not 2008 R2. From everything I have read R2 really should have been almost another version of Windows: Sort of like what Windows 7 (or as some wags call it, Vista Service Pack 3) was to Vista. Inverts the Server 2008 / 2008 R2 naming a bit. What's in a name? I digress....


I have a number. Ancient RHEL 5.1 Linux is 12% more power efficient than MS Windows 2008.


Now my intuition is telling me that by today's Linux release and vendor hardware standards, it is low.