Surprising Power Consumption Of Ubuntu 11.04 vs. Win 7 | Hacker News Reader

Surprising Power Consumption Of Ubuntu 11.04 vs. Win 7

phoronix.com

14 pointsriledhel17 comments

If the power usage really is anywhere near as high as the graph says -- 30 to 45 watts range -- then the laptops aren't going to get much battery life. That is a lot of power to be using.

I suspect that they were testing on the plugged into wall socket power profile, where speed is the main concern, and not power usage. Power consumption changes when the laptop is running off of battery.

Edit: My X200 Tablet runs at about 12 watts on Linux right now, on battery.

They were plugged into mains power, as the first page says:

"For all testing, the Watts Up Pro power meter was used(...) as the power meter was attached to each system's power supply."

That meter measures AC power, which means that for laptop use, as you say, the tests are meaningless.

That site is a piece of shit slow site, can anyone please sum up because by the time I page though all 7 pages I will shoot myself.

I'm really curious as to what the numbers are.

Plus the ads which popup on top of one another.

From page 9 of TFA:

> Ubuntu 11.04 actually performed much better than anticipated. Prior to the tests commencing, the expectation was that Ubuntu 11.04 would lose handedly across all systems to Microsoft Windows 7 Professional Service Pack 1. That was far from the case.

> With similar workloads, for the most part the power consumption is comparable between Ubuntu 11.04 and Windows 7 Pro SP1. The only major differences came during Flash-based HD video playback being more efficient under Windows, power consumption while OpenGL gaming, and in select other areas. Ubuntu / Linux actually has the potential to become more power efficient than Microsoft Windows 7 based upon the close findings from today. Once Active-State Power Management (ASPM) is properly fixed up for Linux, there is still a Linux 2.6.35 kernel power regression, a scheduler power regression, and more. Just yesterday on my Twitter feed, the Phoronix Test Suite and I made a discovery of a possible 8% power savings from an entirely different vector. More to come.

Or in completely tl;dr language: everything is the same except Ubuntu Flash is a pig.

Without wishing to trash their efforts, Linux power management is a mess. Nothing works out of the box and you have to play around with powertop to optimise your kernel configuration to get even reasonable battery life. Add to that, the ACPI implementation is buggy as anything.

This, at least in my case when testing on an Acer 4820T, manifests itself as Windows 7 lasting 6-7 hours on battery (!) with heavy Visual Studio 2010 usage, whereas Ubuntu IDLING lasts about 2.5 hours out of the box and 3.5 after tuning it to bits.

It's ok optimising Linux like this but the whole EXPERIENCE is still horrid.

The problem is mainly with bugs (often intentional) in the ACPI implementations of hardware vendors. While they almost always provide hacky windows drivers, they almost never provide linux patches, or even enough info for correct blacklists.

That's a problem of a failure to code to what is, not to what one thinks it should be.

As I've said before when this comes up--Windows is, for better or for worse, the de facto standard. Alternative operating systems (and Linux is very, very alternative, in this sort of use scenario) that don't recognize and acknowledge the de facto standard's effective supplanting of the de jure standard are Doing It Wrong.

edit: Swapped my jures and factoes. Shouldn't post while writing code...

Did you mean "de facto" when you wrote "de jure", and vice versa? Windows is the de facto standard, and ACPI is the de jure standard; you wrote the opposite.

Er, yes. My bad, was writing a unit test at the same time. ;)

No the problem is that ACPI is a poor piece of design to abstract what is effectively a stinking great turd of an architecture.

You just simply don't get these problems on non-x86 systems. (I speak from experience with MIPS, ARM and SPARC btw).

On non-x86 you get different problems, such as the kernel needing special-case initialization code for every phone model ever made.

That's a fair point. The main problem with that is that most vendors want platform lock in so they implement their own verified loaders (if it's for a phone).

If we had OpenBoot on ARM, things would rock.

So, did you make any powertop suggestions permanent, or are you fiddling with it every time, depending on the situation?

Turned on everything it said would reduce consumption.

Is there a dedicated graphics processor or switchable graphics in this notebook? This might explain the rather big difference in battery time since many Linux graphics drivers are not power-optimized and switchable graphics does not work. [1]

Anecdotally, my rather old notebook (Fujitsu-Siemens Amilo Pro V3205) runs cooler, quiter and longer when using Linux compared to Windows Vista. I still get slightly more than 2 hours out of my battery (3000mAh left). Maybe this is less of a problem with older, well supported hardware.

[1] But people are working on it, see e.g. https://github.com/MrMEEE/bumblebee

Reply on news.ycombinator.com