Ubuntu 9.04 vs Windows 7 Energy consumption test

“The Ubuntu computer uses 15% less energy”


There was some discussions in our office about the consumption of electricity by computers using different OSes.

No real conclusion was given during the conversations and no evidence or tests were found, so my boss decided to run to the electronics store and purchase a couple of electronic meters to once and for all prove what OS saves the most energy.

We only wanted to test relevant OSes and as our company are highly focused on open source software and users of Ubuntu at work, the distrubution of our choice naturally became Ubuntu.
This is also one of the most mainstream distros and thereby very relevant to test.

Of course we needed to compare with a Windows OS as it is the most popular OS family.
The newest contribution to the Windows family (Windows 7) was released as a RC.
This RC became the opponent in our test, and is very relevant as Microsoft states in this blog I have to assume it is the most energy efficient OS from Microsoft up to this date.
Without any further evidence we believe that Microsoft is being honest in this aspect and the older versions of Windows should be less efficient.

Luckily I was assigned this pretty interesting task.
I was to install Windows 7 RC 32 Build 7100 on one of our HP 550 and Ubuntu 9.04 32 on another identical HP 550 and push the computers through different tests while monitoring the meters.
This is the outcome of my tests.

Some of the employees at this office use Macbooks and while it would be very interesting to compare the OSes above with Mac OS X we can’t use identical hardware; the test wouldn’t be fair.
But as a side note the Macbook seems to use less energy than HP 550 with either Ubuntu or Windows.

Technical summary:

The computers used are 2 identical computers of model HP 550 with AC adapter plugged into the electronic meters and with batteries unplugged.
These are the specifications according to the HP website.

Processor type Intel® Core™2 Duo Processor T5670
• 1.80 GHz, 2 MB L2 cache, 800 MHz FSB
Chipset Mobile Intel® GME965
Standard memory 1 x 1024 MB DDR2
Memory slots 2 SODIMM slots supporting dual channel memory
Internal drives 160 GB
Harddisk speed 5400 rpm
Optical drives DVD+/-RW SuperMulti DL LightScribe Drive
Display size 15.4 inches diagonal
Video resolutions 1280 x 800 WXGA BV
Graphic subsystem Intel® Graphics Media Accelerator X3100
External I/O ports 3 USB 2.0 ports, VGA, stereo microphone in, stereo headphone/line out, power connector, RJ-11, RJ-45
Expansion slots Slots available for additional devices: 1 ExpressCard/54 slot, Secure Digital (SD) slot
Audio High Definition Audio, mono speaker, stereo headphone/line out, stereo microphone in
Network interface Intel Fast Ethernet Integrated Controller (10/100 NIC)
Modem 56K modem
Wireless technologies Intel® Wireless LAN 802.11a/b/g, Bluetooth 2.0+
Power requirements External 65W AC adapter, HP Fast Charge Technology
Power features 6-cell (47 WHr) Lithium-Ion battery (Unplugged while testing)

The meters show current Watt consumption and has a timer and a kWh counter.


Ubuntu 9.04 32 bits
Fully updated.
Standard install and no configuration.

Windows 7 RC 32 bits Build 7100.
Fully updated.
Standard install and no configuration.

Applications used during the test are mostly natively included ones.
For example I used Windows Media player 11 vs Movie player and Internet Explorer 8 vs Firefox 3.0.x. Etc.,
although some applications will be identical.


Day 1:

What to test first?
Well, I was in the mode for testing, the first fact is that Ubuntu boots to desktop faster and even though it’s not the purpose for this test I just had to do it.
Ubuntu boots in 53 seconds while Windows 7 boots in 1 minute and 8 seconds.

[iframe src=”http://www.youtube.com/embed/HzVfkwMlo9U” width=”100%” height=”480″]

Now let’s go on topic.
Well both computers are now booted and the meters are running started at the exact same time.
I get about an hour of testing. Before the day is over the meters will measure the total kWh for this hour and I will try it again for a longer period after the weekend.

During this hour I have to come up with some tests.
How about ”idle mode” with no applications running at all.
Ubuntu shows a constant consumption of 33 W while Windows 7 shows a constant 36 W.

Ubuntu is more efficient with nothing running but how about som standard office work.

I have installed Open Office 3.0 on the Windows 7 computer before starting the meters.
While starting up OO the Windows 7 meter peaks at 51 W while Ubuntu only peaks at 41 W but this lowers a bit during the document creation. I see about a 5 W difference in favor of Ubuntu during the short test period.

To do a relevant test on some actual office work 1 hour before the weekend I started downloading the Big Buck Bunny movie. Speeds are about 1.7 MB/s at both computers.
The Ubuntu meters while downloading is a constant 36 W while using Windows 7 got the meter up to a constant 41 W.

Let’s play the movie too using Media player 11 vs Movie player.
The Ubuntu meter is not constant, while playing it changes between 37 and 40 W.
The Windows meter changes between 40 and 42 W.

One hour of testing is not that much but I got some interesting facts.
Ubuntu does seem to be more energy efficient. Before I go home I look at the kWh meter.
It shows 0.04 kWh for Windows 7 and 0.03 kWh for Ubuntu.
I will be testing for a longer period after the weekend.

Day 2:

Time to start the meters again and boot up, I will try to measure as long as possible today.

While figuring out what next to test the computers idle.
After about 15 minutes the energy saving functions start as configured by default.
Both computers turns off the screen but with the Ubuntu meter showing no difference in consumption.
A solid 31 W is showing while Windows 7 dips down to 24 W.

I wait even longer to see what happens.
Windows 7 hibernates itself and while in hibernation mode it only consumes 9 W.
Ubuntu does not hibernate itself but uses only 18 W idling.

Microsoft seems to have better energy saving functions at default setup on Windows 7.
Well thats good, but it isn’t actual office work so on with some actual usage of the computers.

The next test is some Internet activities using Internet Explorer 8 vs Firefox 3.0.x.
I know that the Swedish site www.aftonbladet.se is totally bloated with flash based commercials etc.
Let’s go in there check it out. It must be one of the most demanding site on the Internet.

The meters are pending but again Ubuntu is the winner, by around 10 W.
Let’s push it!
20 tabs of the same demanding site.
The Ubuntu meter varies between 44 and 47 W while Windows is pretty constant at 51 W.

I want to push the computers even more so let’s install Blender 2.49b and render a frame from this project.

While rendering I clearly notice that these computer aren’t really the best choice for rendering.
Both computers kind of lock down and the system monitors show maximum activity on the processors.
No real feedback until about half an hour.
But the meters still show 47 W on Windows and 45 W on Ubuntu.
Not as big a difference as before but still Ubuntu wins it again.

Then I saw the rendering times:
Windows ends at 25 minutes and 29 seconds while
Ubuntu ends at 26 minutes and 34 seconds.

It is possible that it all evens out in this test because of the lesser rendering time for Windows 7.

The last test I am going to write about is yet another ”stress” test.
10 open office documents while surfing 5 tabs of www.aftonbladet.se.
I hit the highest Watt during these test and its Windows 7 again at 54 W.
The Ubuntu meter shows between 44 and 47 W.

I look at the meters.
4 hour of usage and the Windows 7 meter stops at 0.16 kWh
The Ubuntu meter stops at 0.14 kWh.


It’s fairly clear with all these facts, using Ubuntu at work will be more energy efficient.
5 hours of office work and the Ubuntu computer used 0.17 kWh while Windows 7 used a total of 0.20 kWh and this test was made on identical hardware.

The Ubuntu computer uses 15% less energy in 5 hours.
You have to figure out yourself how much money you will save in a year by migrating.