Ubuntu 10.04 Is More Power Hungry Than Windows 7

[kraftman]

Now, none. We agree in the theory.

So let's think for a moment about what it takes for this benchmark to be valid. You mentioned daemons running in Ubuntu, screen brightness settings and lack of power management features in Nouveau. I know that those were only examples, but as such I imagine that you consider them to be representative of the kind of thing that would invalidate the data.

Good to hear this Those were only examples, but I've got your point. I don't have any objections to what you wrote later and I very appreciate you bothered to do so. I usually have objections to the article titles and sometimes to some parts of the article itself. I also believe there are better ways to do the benchmarks, but as far as the defaults are tested it's understandable, but can be misleading.

[fewt]

The whole problem for this test/benchmark is not the high power consumption of Ubuntu 10.04 compared to W7. No, it's the numbers itself.

When the numbers of the netbook are correct, this specific netbook consumes more energy under Ubuntu 10.04 (8W) and only slightly less (4W) when using W7, than a far more powerful Thinkpad T61.

So the question is not, why Ubuntu consumes more energy, but could this numbers be correct.

Both systems use the same version of the proprietary NVdidia-driver, so both systems can use the power saving functions of the used Nvidia graphics-GPU.

An ATOM330-CPUs consumes less than 1W on idle, a C2D T9300 consumes ~8W on idle. The ION-chip aka Geforce 9400 also does consume less energy on idle than the NVIDIA Quadro NVS 140M.

The 12.1" display of the netbook is smaller and has LED-backlight, so the power consumption should be smaller than the 15.4" display of the T61.

So how can it be, that a netbook consumes more (Ubuntu) or slightly less (Windows 7) power than a far more powerful mobile workstation (T61)?

It can't. My T400 uses 14 watts full power save, and my 1000HE netbook uses 6 watts full power save. My T61 used 13 watts, and my T60 used 12 watts. My 1000HD uses 8 watts full power save (Celeron 900).

That was full power save under Ubuntu 9.04 and Eeebuntu v3 with my utilities to manage power. If 10.04 uses that much power, then it is VERY broken or misconfigured out of the box.

[glasen]

The whole problem for this test/benchmark is not the high power consumption of Ubuntu 10.04 compared to W7. No, it's the numbers itself.

When the numbers of the netbook are correct, this specific netbook consumes more energy under Ubuntu 10.04 (8W) and only slightly less (4W) when using W7, than a far more powerful Thinkpad T61.

So the question is not, why Ubuntu consumes more energy, but could this numbers be correct.

Both systems use the same version of the proprietary NVdidia-driver, so both systems can use the power saving functions of the used Nvidia graphics-GPU.

An ATOM330-CPUs consumes less than 1W on idle, a C2D T9300 consumes ~8W on idle. The ION-chip aka Geforce 9400 also does consume less energy on idle than the NVIDIA Quadro NVS 140M.

The 12.1" display of the netbook is smaller and has LED-backlight, so the power consumption should be smaller than the 15.4" display of the T61.

So how can it be, that a netbook consumes more (Ubuntu) or slightly less (Windows 7) power than a far more powerful mobile workstation (T61)?

[yotambien]

I meant data means nothing if the procedure was screwed or if the procedure is not known. You said "you can't blindly accept a bunch of numbers without looking at them.", so what's the problem?

Now, none. We agree in the theory.

So let's think for a moment about what it takes for this benchmark to be valid. You mentioned daemons running in Ubuntu, screen brightness settings and lack of power management features in Nouveau. I know that those were only examples, but as such I imagine that you consider them to be representative of the kind of thing that would invalidate the data.

Daemons running in Ubuntu. This is an acceptable objection if we are talking about an updatedb run or something similarly intrusive and discrete in time. It wouldn't make sense to check the power consumption while a process like this eats up loads of CPU cycles and trashes the hard drive. On the other hand, daemons that run continually and are enabled by default should not be excluded, in the same way that nobody would complain about leaving intact Windows services. In any case, the article explicitly says "we simply monitored the power consumption [...] when each operating system was idling at their respective desktops with all default settings and software left for each OS".

Screen brightness. This is not a valid complain. There's no reason to believe that Michel touched the brightness settings--quite the opposite--so basically it was left up to the OSs to control it if needed to. If Ubuntu or Windows have different default values or different screen-related power saving policies, that has to be taken into account. There's no doubt you can minimise the power consumption in both OSs by tweaking this and that. It also has been said to death that the idea is to benchmark the systems at their default values.

Lack of power management in Nouveau. Not valid. As in not at all. If this driver doesn't have this feature that's just too bad, and the data rightly and conclusively shows it. A perfect example indeed. On top of that, care has been taken to show the numbers from both the freshly installed and updated systems, so you can compare what effect another driver has.

[kraftman]

Groarr, you are not paying attention or you just want to have an argument out of nothing. I know--really well--that you can't blindly accept a bunch of numbers without looking at them.

Not knowing what exactly were settings and conditions it's like accepting the data and not looking at it.

But once you ascertain their validity and understand their scope, they become useful data from which to extract conclusions about whatever it is they refer to.

Yes, but only if you're able to do so.

You just presented an imaginary screwed-up scenario that, in any case, doesn't support your previous assertion about "data itself [meaning] usually nothing".

If some daemon was running in Ubuntu, screen was brighter etc. it's not an imaginary scenario. While data collection procedure was screwed-up such data means nothing. Nothing important and nothing interesting. If nobody cares about the procedure then such data is usually meaningless, because it shows some numbers, but they're not related to each other. "Ubuntu is more power hungry then Windows 7" - data in the article does not prove this. It only shows the Ubuntu was consuming more power using proprietary nVidia driver and Nouveau driver which doesn't support Power Management. We don't even know if it's screen was brighter or not.

So again, what you are saying--hopefully--is that the data collection procedure has to be adequate. We all agree about that. Note that I'm not saying a word about this particular benchmark; but I couldn't resist making a comment about that outrageous sentence of yours.

I meant data means nothing if the procedure was screwed or if the procedure is not known. You said "you can't blindly accept a bunch of numbers without looking at them.", so what's the problem?

You refer to something like this, unless the numbers have drastically changed since some months ago:

Linux: 47.4%
Windows: 41.3%
Macintosh: 7.9%
Unknown: 2.6%
BSD: 0.2%
Solaris: 0.1%

But you have to realise that many people access Phoronix from work with a Windows machine.

Yes, my mistake, but it doesn't matter.

[Messerjocke]

Good powermanagement at least for my laptop platform

Seems I am lucky with my hardware:

A lenovo G560 laptop with discrete NVidia 310m graphics and Core i3 CPU.

It idles at around 13 Watts.

The GPU and its memory get clocked down to 133MHz in desktop mode (nvidia driver default settings) while the CPU snores at 933MHz (ondemand).

I get the same results and battery runtime under Windows 7 using all the clever Lenovo-Powermanagement stuff.

[stevea]

but when both were loaded with NVIDIA's binary driver that leverages PowerMizer and other power-savings techniques, Ubuntu 10.04 LTS averaged to consume just 9% more power.

Since when is 25watts vs 24watts "9% more" ?

[yotambien]

Nonsense. Data, as opposed to hearsay and handwaving, means everything.

When one system is theoretically updating the database and it's compared against another one which is drinking a coffee such data means nothing. You'll have numbers and you can then try to analyze everything, but an only data won't tell you what those OS's were doing, so your conclusion can be simply wrong.

Groarr, you are not paying attention or you just want to have an argument out of nothing. I know--really well--that you can't blindly accept a bunch of numbers without looking at them. But once you ascertain their validity and understand their scope, they become useful data from which to extract conclusions about whatever it is they refer to.

You just presented an imaginary screwed-up scenario that, in any case, doesn't support your previous assertion about "data itself [meaning] usually nothing". So again, what you are saying--hopefully--is that the data collection procedure has to be adequate. We all agree about that. Note that I'm not saying a word about this particular benchmark; but I couldn't resist making a comment about that outrageous sentence of yours.

The whole point of the Phoronix test suite is to collect data, which can afterwards be analysed and discussed. I guess you meant to highlight the importance of the interpretation of the data, or perhaps its validation. If that is the case I concede the point, but I'm just speculating about what you wanted to say...

Then the whole Phoronix point is dumb stupid. They should care about the as much ideal situations as possible, so the data numbers will be much more interesting and meaningful (and to not let some random situations happen like doing something in the background or similar). In your opinion they're forcing readers to analyze provided data, so they're also misleading others (who do not know they should do this) by providing some kind of data and using misleading article titles? While the data must be analyzed why there are made some assumptions just after the benchmarks?

You are mixing things up, making up others, and overall missing the point. It is not for me to sort that out. But I can tell you that I personally enjoy the forum discussions that follow the benchmark articles. However flawed some of the benchmarks sometimes may seem, there is almost always something to learn from them and the comments of some readers.

Afaik there was some New year's sum up and there were more Windows users visitors.

You refer to something like this, unless the numbers have drastically changed since some months ago:

Linux: 47.4%
Windows: 41.3%
Macintosh: 7.9%
Unknown: 2.6%
BSD: 0.2%
Solaris: 0.1%

But you have to realise that many people access Phoronix from work with a Windows machine.

[kraftman]

but an only data won't tell you what those OS's were doing,

The same data...

[kraftman]

Walk up to over whelming majority of laptop owners and ask them how does their system provide good battery life. They don't know and most of them don't care. All they care about is the amount of battery life they do get, not how it is achieved.

If only testing conditions were as they should be it's ok, because if one screen was lighter or Ubuntu on netbooks has some additional package installed which helps it to operate longer on batteries such comparison is meaningless. However, the tittle of the article is still very misleading.