test: OS power consumption

Ecological issues around computing. This is an experimental forum.

Moderators: Ralf Hutter, Lawrence Lee

Post Reply
Palindroman
Posts: 247
Joined: Fri Oct 20, 2006 1:00 am

test: OS power consumption

Post by Palindroman » Sat Jan 26, 2008 11:43 am

A few days ago I tested various operating systems to see what their power consumption would be on exactly the same setup:

Gigabyte GA-M68SM-S2 (nVidia 7025/630)
AMD Sempron LE-1100
1 x 1024 MB RAM, 667 MHz, ECC
1 Samsung HM080JI (2.5")
picoPSU-120 with 80W Morex Adapter
Samsung SH-S203B dvd-rw

I tested four OSs: XP, Vista, Ubuntu 7.10 and gOS:

Image

I read about this but wasn't sure if it was true: Vista doens't seem to consume a whole lot more tha XP. Added to this is the fact that in Vista Cool 'n Quiet seems to be turned on automatically (didn't know this). The system could handle Aero with a score of 3.0.

Ubuntu consumes just as much power but unfortunately it doesn't have its own version of RMClock yet. The last OS I tested is an Ubuntu version that is sold with the $200 Everex gPC at Walmart. Many people seem to think the 'g' in gOS stands for 'green', but it stands for 'good'. As you can see it's not very 'green' on this setup but that's also because it was designed specifically for VIA. The gPC however is said to consume 40 watts at idle which makes the whole scheme sound like gWash.

Any thoughts on this? And is there any other OS someone would like to see tested on this setup? Gentoo or some other distro perhaps?

floffe
Posts: 497
Joined: Mon May 08, 2006 4:36 am
Location: Linköping, Sweden

Post by floffe » Sat Jan 26, 2008 12:02 pm

Maybe you could try with the still rather untested kernel patches from here, the centrino version works good with my pentium M laptop and I see that they recently added stuff for amd cpus. It requires building a custom kernel, possibly with these patches checked out from svn, so it'll be rather involved. Gentoo would probably be a good idea since they stay rather out of the way for making your own kernel, and I could help you there but I'm not too sure about how to customise it on ubuntu.

And I have a question: Was Vista running Aero for this test? Did ubuntu have the 3d effects enabled? Did the numbers change if you toggled those settings?

Palindroman
Posts: 247
Joined: Fri Oct 20, 2006 1:00 am

Post by Palindroman » Sat Jan 26, 2008 12:19 pm

floffe wrote:Maybe you could try with the still rather untested kernel patches from here, the centrino version works good with my pentium M laptop and I see that they recently added stuff for amd cpus. It requires building a custom kernel, possibly with these patches checked out from svn, so it'll be rather involved. Gentoo would probably be a good idea since they stay rather out of the way for making your own kernel, and I could help you there but I'm not too sure about how to customise it on ubuntu.
I talked to these guys and they gave me a patch but to be honest I don't have the faintest clue what to do with it. I only know a few things about computer hardware but none whatsoever about software. When it comes to Ubuntu all I know is how to use the terminal and follow step-by-step Howto's. If you could help me along I'd be very grateful. I always get the hang of things but it takes a while. :)
And I have a question: Was Vista running Aero for this test? Did ubuntu have the 3d effects enabled? Did the numbers change if you toggled those settings?
Yes, Aero was runnign and it didn't make any noticeable difference. I only found out the 3d effects in Ubuntu today (that's Compiz, right?) but haven't really paid any attention to power consumption. Will do that after the weekend when I have time to do some more testing.

Michael Sandstrom
Posts: 606
Joined: Wed Oct 13, 2004 4:03 pm
Location: Albany, GA USA

Post by Michael Sandstrom » Sat Jan 26, 2008 4:02 pm

Although this power consumption comparison is interesting, readers should realize that these tests don't take into consideration the differing hardware requirements of each OS. Vista's hardware requirements are much more stringent than Windows XP's which are much more stringent that Ubuntu's.

floffe
Posts: 497
Joined: Mon May 08, 2006 4:36 am
Location: Linköping, Sweden

Post by floffe » Sat Jan 26, 2008 4:35 pm

Palindroman wrote:I talked to these guys and they gave me a patch but to be honest I don't have the faintest clue what to do with it. I only know a few things about computer hardware but none whatsoever about software. When it comes to Ubuntu all I know is how to use the terminal and follow step-by-step Howto's. If you could help me along I'd be very grateful. I always get the hang of things but it takes a while. :)
As I said, I'm not too sure about Ubuntu in particular, but the main custom kernel howto for that is here, although that doesn't seem to include the step where you actually apply the patch. There's also a howto at a thinkpad forum, which I think you can pretty much follow as long as you make sure to apply the patch you got instead of checking it out from svn (the patch -p1... step), and build the powernow-k8 instead of speedstep-centrino.

And yeah, compiz takes care of visual effects in ubuntu.
Last edited by floffe on Sun Jan 27, 2008 2:01 am, edited 1 time in total.

drees
Posts: 157
Joined: Thu Jul 13, 2006 10:59 pm

Post by drees » Sat Jan 26, 2008 11:02 pm

When testing Linux, have a look at the tool PowerTOP to help narrow down causes of un-needed power consumption due to processes waking up too often.

NoMoreNoise
Posts: 51
Joined: Wed Oct 20, 2004 3:16 am
Location: Norway

Post by NoMoreNoise » Mon Feb 04, 2008 10:40 am

What about the fact that the Linux distro is probably a lot less energy consuming to develop than MS OS ?

All those people working for MS has go to get to work somehow....

Palindroman
Posts: 247
Joined: Fri Oct 20, 2006 1:00 am

Post by Palindroman » Sat Feb 09, 2008 3:14 am

SPCR forum member aaa is helping me to get PHCtool to work on Ubuntu which means I'll hopefully be undervolting my Sempron soon. I'm really curious what the power consumption will be compared to XP and Vista. I'll post here as soon as I find out.

Let's suppose I get this to work and everything runs stable. What happens when I upgrade Ubuntu to a newer version. Will the CPU remain undervolted?
drees wrote:When testing Linux, have a look at the tool PowerTOP to help narrow down causes of un-needed power consumption due to processes waking up too often.
aaa told me that all this is incorporated in Ubuntu 7.10. However I do notice that the hard disk never seems to go into some kind of idle/sleep mode, or only for a very short period like 1 or 2 seconds. In XP the HD after a minute or so becomes quieter and you see power consumption drop an extra half a watt.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Thu Nov 27, 2008 4:55 am

Given the very same hardware I see no reason why there should be any difference whatsoever in power consumption during idle.
- There is no computing going on.
- There's no reading or writing to the HDD.
- The screen shouldn't change.

To figure out which OS is more efficient it's better to see how much energy (power integrated over time) is used to solve a given series of tasks, from "go" to finished.

Unfortunately this is a type of test that's very difficult to perform in a repeatable and objective manner.

Cheers
Olle

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Post by CA_Steve » Thu Nov 27, 2008 8:03 am

Supposedly, the big diff in XP vs Aero power usage is that Aero Glass makes the GPU run in 3D mode instead of 2D as in XP. So, if your GPU consumes little power when idling in 3D, that could explain the results. <shrugs>

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Thu Nov 27, 2008 12:27 pm

Olle P wrote:Given the very same hardware I see no reason why there should be any difference whatsoever in power consumption during idle.
- There is no computing going on.
Except that different OSes enable different kinds of background services by default, which still do stuff when the computer is otherwise idle.

Most of these the knowledgeable user can configure away.
Olle P wrote: - There's no reading or writing to the HDD.
Windows definitely seems to be accessing disks pretty much all the time. Again this has to do with the services enabled, logging, file system metadata etc.
Olle P wrote: - The screen shouldn't change.
If the screen is properly configured to set a blank screen and monitor power off at idle. But if the user set some "screensaver" then this is not the case.
Olle P wrote: To figure out which OS is more efficient it's better to see how much energy (power integrated over time) is used to solve a given series of tasks, from "go" to finished.

Unfortunately this is a type of test that's very difficult to perform in a repeatable and objective manner.

Cheers
Olle
For typical home use tasks, there's probably not much difference. These start to matter with commercial server farms doing some heavy number crunching.

However I do agree with you on your original point if we are comparing different Linux distributions against each other. There should be no point in comparing the power usage of different linux distros, because you can have your kernel compiled with exactly the same options on each one, and enable/disable exactly the same services. Pretty much all the software is available for each distribution, there's just a different way to install it on each one.

drees
Posts: 157
Joined: Thu Jul 13, 2006 10:59 pm

Post by drees » Wed Dec 03, 2008 1:38 am

lm wrote:
Olle P wrote:Given the very same hardware I see no reason why there should be any difference whatsoever in power consumption during idle.
- There is no computing going on.
Except that different OSes enable different kinds of background services by default, which still do stuff when the computer is otherwise idle.
Even more than that - it's up to the device drivers to put devices into low-power modes and not all device drivers do that by default or even make the option possible.

Even when a computer is "idle", you'd be surprised how much is actually still going on and how often the CPU is woken up to to meaningless tasks. Each time the CPU is woken up to do one of those tasks, power consumption will go up.

Have a look at the Linux Powertop project for more insight as to how various items can affect "idle" power consumption.

Post Reply