Ubuntu 10.04 Is More Power Hungry Than Windows 7

[Ragas]

Changing the fan is not an option (I tested), because with ondemand core temperature simply rises to the point where the fan triggers again (tried up to 64 degrees - anything over that is uncomfortable, both physically (it burns!) and conceptually (nvidia bumpgate, better not push it)).

I once had a Problem like yours where the laptop reached the spinning up temperature on high clocks, afte some amount of time it cooled down a little and sun down ... in the end it left me with a constantly up- and down-spinning system, which was really annoying. and there it helped me to increase the upspinning value a tiny bit to have it silent. which is why I suggested it could help.

Right now I'm using PHC so actually it never gets that hot anymore. cause the processor does maximum clock at minimum voltage.

[krazy]

A couple of points:
- Proprietary drivers are crappy. Please try this with intel integrated graphics.
- Please install laptop-mode-tools. The kernel has a laptop-mode which isn't enabled without this package.
- Please retest using battery-power, as this is a much more relevant and important test. Some power-saving techniques are only enabled when on battery power.

[locovaca]

I find it curious that you're using ubuntu to do any performance tests. I've found it to be really very slow in relation to the hardware performance, which also translates into higher power use (although that's pretty much par for the course for all the newer distributions).

You may want to check out alot of the other articles on this site with the head to head comparisons. It's fair to say that Ubuntu is at least on par with other distributions, but usually exceeds others in performance.


As to the power tests, this isn't your run of the mill netbook. It is a Dual Core Atom (4 watts right there) with an Ion chipset (even more power draw). I am not surprised by the results. If you read other articles citing power numbers for the Ion platform you'll see that they usually found low to mid 20's on DESKTOP systems. Take away the desktop hard drive and replace it with a LED LCD + Laptop Hard Drive and the results seem to be pretty appropriate.

[notzed]

I find it curious that you're using ubuntu to do any performance tests. I've found it to be really very slow in relation to the hardware performance, which also translates into higher power use (although that's pretty much par for the course for all the newer distributions).

[mugginz]

While strictly speaking the title of the article is accurate, it looks somewhat unnecessarily inflammatory

Perhaps the title should read "Ubuntu 10.04 Is More Power Hungry Than Windows 7 while on AC power"

[mugginz]

A netbook that consumes as much energy as a far more powerful mobile workstation aka Thinkpad T61, is not plausible.

The results certainly do surprise.

Given that the machines were plugged into AC power, surely all that the tests are measuring are a given operating systems power consumption while connected to an AC outlet. This in itself has diminished value when one wants to make decisions based on power usage while not connected to AC, and instead running solely off batteries.

If someone wants to make operating selection decisions that revolve around how much on battery usage time is available, it would seem they should really wait for further tests. While strictly speaking the title of the article is accurate, it looks somewhat unnecessarily inflammatory

Numbers are generated and are never checked for plausibility. And if there is prove that the numbers are incorrect, they will never be corrected and the flawed article stays online forever.

It looks like the methodology isn't quite what's needed in order to draw conclusions for what someone can expect while away from an AC outlet.

I do think it's valid to use Ubuntu to represent Linux in the tests given its popularity and is equaly valid to test it in its vanilla configuration as this is what a lot of people will use. There would be additional value though in including results for a fine tuned configuration in order to provide a clearer picture for what someone might expect while using a system that's been set up by a more expert technician.

[glasen]

Wild guess: was the battery connected during those tests? Maybe it was recharging at the time, invalidating the results?

This could be the answer.

Sorry for the offense, but this test is typical for Phoronix :

A netbook that consumes as much energy as a far more powerful mobile workstation aka Thinkpad T61, is not plausible.

Numbers are generated and are never checked for plausibility. And if there is prove that the numbers are incorrect, they will never be corrected and the flawed article stays online forever.

[0laf]

Hi Micheal,

the numbers of your energy meter really seem a bit off, for example take a look at [1].

[1http://www.notebookcheck.net/Review-...k.24498.0.html]

[BlackStar]

sure. but bgamari is right either way. the philosophy is to ramp up the processor fast and do as much work as possible to be able to sleep longer after that, wich saves much more power, than running the processor slowly.
It's like in real life: you get your work done fast to have more time afterwards to do what ever you want (save power).

that your fan ramps up more is possible (more heat is generated in short time, which doesn't mean that this is true in overall time) ... if it disturbs you, you could change the spinning up and down values a little to avoid that.

As I said I'm not interested in power consumption only in fan noise. Ondemand causes the core temperature to burst to 60+ degrees when scrolling e.g. Firefox, triggering the fan. Conservative keeps the core temperature to 57-58 degrees, which is silent.

Changing the fan is not an option (I tested), because with ondemand core temperature simply rises to the point where the fan triggers again (tried up to 64 degrees - anything over that is uncomfortable, both physically (it burns!) and conceptually (nvidia bumpgate, better not push it)).

In other words, I don't really care if ondemand would save me 1WHr on average: I'm only interested in silent operation.

Btw, I find the results of this article very strange. My D830 (65nm Core 2 mobile 1.8GHz, Quadro NVS135M, 15.4'' 1680x1050 S-PVA monitor) consumes 27W on idle at full brightness (both Windows and Ubuntu). My Atom netbook (Lenovo S10-2) consumes 12W on idle on Ubuntu and 11W on Windows. 25W for a netbook is way, way too high.

Wild guess: was the battery connected during those tests? Maybe it was recharging at the time, invalidating the results?

[yotambien]

Mr. Garrett is wrong. Reducing the clock speed reduces power output which in addition to lowering the amount of power used per clock cycle, it also allows your BIOS to reduce FAN speed due to reduced heat. Intel themselves (who wrote the powertop utility) recommend down clocking a CPU to further reduce power output. "C" states ABSOLUTELY reduce power consumption by idling functions within the processor essentially shutting down a core if you will, but reducing the clock rate in addition lowers power utilization even further.

M. Garrett acknowledged in a footnote of that blog entry that things are not as clear cut as he put them in the main text. A bit annoying, yes.

The rule of 'do everything as fast and possible and iddle as much as possible' seems to be an oversimplification only valid when compared to the performance governor. However, I remember seeing some CPU benchmarks where the performance and power consumption was measured; they found that the more power hungry CPUs in the iddle state were much more power efficient under heavy load.

Do note that Ubuntu boots into performance and reverts to ondemand after 60 seconds or so. This is done to improve boot times and does impact power consumption.

(I disabled this feature today in favor of using the "conservative" governor everywhere. My mobile Core 2 doesn't really need the extra performance to boot).

I guess during the initial booting sequence the cpu frequency is set up by the BIOS, but as soon as the system is capable of doing so it should really set the 'ondemand' governor. It makes no sense that the CPU is all revved up while, say, waiting for a network share to be mounted or whatever. In any case, the work aimed at tweaking the boot time in Ubuntu probably minimises the impact of this policy.

I too hate the laptop fan. It's not particularly noisy, but it makes no sense that it has to kick in while scrolling a website or pdf document. To avoid this I used to limit the maximum CPU frequency, but that's suboptimal and requires manual intervention. However, it was the only way I had to avoid hearing the fan while playing some flash games or even MAME, which don't need all the cycles my CPU is capable of. I made some trials using the conservative governor and I'm sold. For starters, it doesn't go straight away to the highest frequency, but sequentially increases it until it doesn't reach the value specified in some tunable parameter.

I highly recommend the following article, which not only explains the different power saving policies and options you can set up, but also shows benchmarks and explains how to slightly tweak them to maximise power efficiency while keeping performance almost intact and latency to a minimum:

http://www.ibm.com/developerworks/linux/library/l-cpufreq-1/