Ubuntu's Power Consumption Tested

[phoronix]

Phoronix: Ubuntu's Power Consumption Tested

For this article we've decided to not only deliver power benchmarks from Ubuntu 7.10 and Ubuntu 7.04 to compare the tickless kernel effect, but we have went back and retested all of the Ubuntu releases going back to Ubuntu 5.04, or also known as Hoary Hedgehog. With the past six Ubuntu releases we had tested the power consumption of a Lenovo laptop when running from its AC charger and off the battery, when the system was idling and then again under load. We had also monitored the temperature of the Intel Centrino mobile processor. You may be surprised by the results of Ubuntu's power usage.

http://www.phoronix.com/vr.php?view=11239

[Regenwald]

thank you again for this really great article! it is exciting to read articles about this topic.
well, it is very sadly that they are not able to reduce ubuntu's power consumption.
perhaps they're going to get it with upstart. i recently read a great article about its future and it seems very exciting (http://liquidat.wordpress.com/2007/09/12/discussion-around-upstarts-future/)
in short: less services/deamons -> less power consumption (why do you need a cupsd when you don't have a printer?)
but i would like to see a test with the current fedora 8. they have optimised the whole system for less power consumption. i would say that it had less power consumption than ubuntu/other distris...

[ssam]

interesting.

it would be good to see similar tests for a range of hardware. judging by intel's recent involvement with linux powersaving, i'd expect thier new chips to benifit most.

does phoronix have a core 2 duo to test? (will old ubuntus boot on a modern chip?)

[Michael]

does phoronix have a core 2 duo to test? (will old ubuntus boot on a modern chip?)

Of course :) We have tons of hardware at our disposal. Old Ubuntus should be able to boot to a Core 2 Duo, but motherboard compatibility is usually a different story.

[mile]

powertop is nice utility

http://www.linuxpowertop.org/

[mintcoffee]

Great article!!

I noticed you mentioned that you used PowerTOP to measure the power usage. PowerTOP can also display the processes with the most CPU wake-ups, which gives us a good idea which processes are consuming more power.

Of course, this requires you to compile your own kernel.

[TinGrin87]

i'd be curious to see the power usage difference between a win OS (say, XP or Vista) vs Ubuntu- even if we could see just 1 or 2 benchmarks vs ubuntu, it'd show that open source actually does have better programmers than some of the big guys, like M$ :)

(please excuse the dreadfulness of this post, it's midnight and i'm tired)

oh and welcome from Digg :)

[narf]

I am curious about your powertop results. What processes are causing interrupts? On my laptop, the radeon driver is generating a 60Hz interrupt. There are also other problems, but the radeon driver is the worst offender.

To me, the most interesting benchmark for a laptop is how long the battery lasts while playing a DVD.

From what I understand, the bset thing to reduce power consumption is to get rid of all the periodic interrupts so that the processor can go to sleep when the machine is idle.

[narf]

mintcoffee, I am running powertop on my laptop with a stock Fedora kernel, and I can see all that process interrupt stuff just fine.

[Prosthetic Head]

The firts thing i'd turn off is tracker as it keeps my hdd scanning all the time. If you ask me it should never have been included at this stage until it could be made less demanding.. Its very IO intensive and also eats the cpu a bit so i'd guess theres quite some power going there.. unfortunatly i don't have a meter to test for my self..

ps: Hi everyone :)

[Ryzzen]

Hmmm, so this was done with just a default installation of Gutsy, yes? By default, it has Tracker (which indexes your hard drive) and Compiz (pretty things) enabled by default, which no doubt have some sort of impact on the results. I'd be very interested in seeing the results with those features disabled, since those would be the first things I'd turn off after installing it on a laptop.

I have a hunch that the power usage would be a whole lot lower. I find it hard to believe that the tickless kernel really has such a small impact...

[ikaruga]

Great article. Though I'm not really surprised by the results. My nonscientific "experiment" of watching the same programs that ran fine Edgy run slowly on Fiesty led me to the conclusion that Ubuntu has become a BLOATED POWER-HOGGING monster.

The article would be greatly improved if you compared various distros (not just ubuntu) and even Windows... I'd really like to see how Vista compares to 'buntu and OpenSuse... I have a feeling that Ubuntu is not too far behind Vista...

[Michael]

The article would be greatly improved if you compared various distros (not just ubuntu) and even Windows... I'd really like to see how Vista compares to 'buntu and OpenSuse... I have a feeling that Ubuntu is not too far behind Vista...

We will be delivering these benchmarks shortly.

[Prosthetic Head]

We will be delivering these benchmarks shortly.
Cool, thanks :)
Can you try ubuntu with tracker and compiz turned off too if you have time while you're doing it?

[ssam]

The article would be greatly improved if you compared various distros (not just ubuntu) and even Windows... I'd really like to see how Vista compares to 'buntu and OpenSuse... I have a feeling that Ubuntu is not too far behind Vista...

and fedora. with their OLPC involvement I think they ought to be ahead on powersaving. i saw a redhat guy at lugradio who had some clever ideas, eg smart screen dimming; the dim timeout varies automatically, if the screen dims and you wiggle the mouse it increases the time out, if you keep leaving the screen dim for a fews mins, then it increases the time out.

also could you check how gutsy with tracker and networkmanager disabled does. the current results might just show that you can run more background tasks without increasing power consumption.

[Luis]

Yes, as reported by powertop, while the system is idle trackerd is in the top 5 offenders, but just as HAL polling the DC/DVD drive to see if you have inserted a disk or like USB port.

I don't think that power consumption while idle is all that important. Probably like 5% of your battery is used by an idle system (excepting the battery used by the screen if it's on), while 95% is used when things are actually happening.

And yes, Tracker again uses quite some power. Not by waking up the kernel while idle, but while scanning the disk and updating the database.

I'd find interesting to measure the power used by applications. For example, it could be measured how much power it's used by 3 different bittorrent clients downloading the same file: Azureus (a full-featured Java based app), Ktorrent (a full featured QT/C++ based app) and Transmission (a simple and light GTK/C app). That would give an idea about the relationship between bloat/features and power consumption.

It's said that why care to optimize apps when more powerful hardware (more CPU, more RAM, etc...) is cheaply available each year. Well, power consumption could be a good reason for it.

[exilist]

Great article. A second one about Power consumption on desktop systems would be nice!