Ubuntu 10.04 Is More Power Hungry Than Windows 7

[Michael]

To those wondering if the numbers are off, no they are not off. In fact, testing under each case was done twice.

[BlackStar]

You should be aware that you are actually crippling[1] the ability of your processor to save power by forcing a lower clockrate. Unless you are worried about your thermal envelope, there is absolutely no reason to deviate from the ondemand governor.

Modern processors save the most power when they have long periods of inactivity as they can then spend longer in deeper C-states. By slowing the processor down, you are requiring that it take longer to complete its work and therefore spend more time dissipating more power.

[1] http://mjg59.livejournal.com/88608.html

Don't wish to derail this thread, but my ears disagree with this assessment. Ondemand is quite trigger-happy and causes the fan to ramp up more frequently than conservative. Scrolling a webpage shouldn't require more than 800MHz(*) in most cases - and with conservative it doesn't. Move to ondemand and the CPU will ramp up to 1.2GHz while scrolling, which is wholly unecessary and thoroughly annoying (fan noise).

(*) on a mobile Core 2.

[HokTar]

So far everybody has been arguing over GPUs.

May other concern, however, is that my CPU is supposed to have 8 power states. Under Linux I only have 3, 550MHz, 1100MHz and 2200MHz (AMD Turion Ultra X2). Knowing that power consumption is proportional to frequency I would love to see a ~275MHz option. Others are welcome, too.

Point is, how is it possible that AMD doesn't show all available options?

[fewt]

You should be aware that you are actually crippling[1] the ability of your processor to save power by forcing a lower clockrate. Unless you are worried about your thermal envelope, there is absolutely no reason to deviate from the ondemand governor.

Modern processors save the most power when they have long periods of inactivity as they can then spend longer in deeper C-states. By slowing the processor down, you are requiring that it take longer to complete its work and therefore spend more time dissipating more power.

[1] http://mjg59.livejournal.com/88608.html

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.

I can reduce my Eee PC 1000HE to *6 watts* utilizing a combination of CPUFreq, SHE (bus downclock), and C states. Working with the Eeebuntu team, I integrated this technology into Eeebuntu 3 and EB4 with great success.

The core flaw with his logic is that you are performing more operations faster, but if you lower the clock rate you perform those same operations at a slower rate but at a reduced power cost.

For the 1000HE it means the difference between playing video non-stop for 5 hours without down clocking, and playing video non-stop for 9.5 hours with down clocking. I have similar results with the Celeron CPU used in the 1000HD and countless others as reported by our users.

http://software.intel.com/en-us/blogs/2008/03/12/c-states-and-p-states-are-very-different/

[cklein]

To those wondering if the numbers are off, no they are not off. In fact, testing under each case was done twice.

Michael, could you please take a look at two issues about how the numbers are presented in the article:

1. Writing 25.00 suggests that you did the measurements with an instrument which has a resolution of 0.01. Seeing that all numbers end in .00, I suppose that this is not the case.
2. I don't understand how 24.00W and 25.00W (especially with .00) lead to the conclusion that Ubuntu is 9% more power hungry.

[FunkyRider]

Ubuntu in default does very bad on power saving, I have to manually tweak a lot of places in 9.10 for it to achieve 9W idle power. I don't know why those laptop related features all get disabled in laptop mode, what are they thinking?

[ebird]

This comparison probably has sense, because it seems out of the box Windows 7 will operate longer on batteries. However, if Ubuntu favors performance over power consumption it would be nice if someone would just mention this.

Yes it is. But this test show only, that a bad supported device consumes more power under linux. A good suported device consumes the same. My Athlon II X3 435 consumes 40Watt (linux) instead of 50Watt in idle mode, without phc and with compiz, because windows 7 is allways doing something in the background. Same with my 4050e home server, but not as bad with Windows XP, 30Watt (linux) instead of 35Watt. The second difference should be just measuring tolerance. On my old Thinkpad R50p I got 3h (linux) instead of 2h battery runtime with a tweaked phc kernel. All my results are useless, a there are different service and programs running the background, not comparable.

[fewt]

Yes it is. But this test show only, that a bad supported device consumes more power under linux. A good suported device consumes the same. My Athlon II X3 435 consumes 40Watt (linux) instead of 50Watt in idle mode, without phc and with compiz, because windows 7 is allways doing something in the background. Same with my 4050e home server, but not as bad with Windows XP, 30Watt (linux) instead of 35Watt. The second difference should be just measuring tolerance. On my old Thinkpad R50p I got 3h (linux) instead of 2h battery runtime with a tweaked phc kernel. All my results are useless, a there are different service and programs running the background, not comparable.

It does not show that at ALL. Eee PCs are very well supported devices.

[lavinog]

I did a similar test a couple of days ago with my older laptop using the open source radeon driver in ubuntu and WinXP.
I find it odd that my older bulkier laptop consumes less power than a netbook.

http://ubuntuforums.org/showthread.php?t=1469921

                                                   Power Consumption (in watts)
                                        WinXP           Ubuntu (Lucid Kernel)   Ubuntu (2.6.34)
Boot                                    29-32           27-50                   29-53
Idle Desktop / Wifi On                  18              19                      18
Idle Desktop / Wifi Off                 15              18                      17
Lid Closed / Wifi On                    15              12.1                    12.3
Lid Closed / Wifi Off                   13              11.3                    11.1
Hd Sleep / Lid Open / Wifi Off          -               16.1                    15.0
Hd Sleep / Lid Closed / Wifi Off        -               9                       9

[glasen]

To those wondering if the numbers are off, no they are not off. In fact, testing under each case was done twice.

Sorry, but than your power-meter does not work correctly with this netbook battery-charger. Why? Because you've already shown that this specific EeePC only consumes 16-20W.

http://www.phoronix.com/scan.php?page=article&item=asus_eee_1201n&num=5

And in this test the CPU wasn't idling!