Mobile Users Beware: Linux Has Major Power Regression

[evolution]

Still out of luck...!

CONFIG_CPU_FREQ_STAT and CONFIG_X86_MSR

Well, I enabled those options as modules in the kernel configuration, but I'm still out of luck with powertop 1.97...

I'm sending a link (pastebin) with my lsmod output and my kernel configuration, if there's something more missing in the kernel configuration...

Cheers

[evolution]

Solved!

Well, I enabled those options as modules in the kernel configuration, but I'm still out of luck with powertop 1.97...

I'm sending a link (pastebin) with my lsmod output and my kernel configuration, if there's something more missing in the kernel configuration...

Cheers

Well, I finally managed to make powertop 1.97 (aka 2 beta 1) work in my 2.6.38 kernel by enabling "default tracers" and "power event tracing" (together with the kernel options you told me before).

Well, the same test I did above with PowerTop 1.13 in kernel 2.6.38 is giving me now (with PowerTop 1.97) a much smaller amount of "wakeups per second" and much less "timer events"...! (screenshot below, click on it to see it in full-size)
My question now is this: How can this be possible?!

Powertop 1.97, linux kernel 2.6.38.4:


Cheers, and btw, thanks for your suggestions fewt!

[fewt]

Well, I finally managed to make powertop 1.97 (aka 2 beta 1) work in my 2.6.38 kernel by enabling "default tracers" and "power event tracing" (together with the kernel options you told me before).

Well, the same test I did above with PowerTop 1.13 in kernel 2.6.38 is giving me now (with PowerTop 1.97) a much smaller amount of "wakeups per second" and much less "timer events"...! (screenshot below, click on it to see it in full-size)
My question now is this: How can this be possible?!

Powertop 1.97, linux kernel 2.6.38.4:


Cheers, and btw, thanks for your suggestions fewt!

Probably because the problem isn't with the kernel, it's with the reporting / benchmarking tool. You seem to have independently confirmed what I have been saying all along.

[Michael]

Probably because the problem isn't with the kernel, it's with the reporting / benchmarking tool. You seem to have independently confirmed what I have been saying all along.

PTS doesn't do anything with measuring wake ups, etc. All that PTS is doing is monitoring the sysfs interface for the power usage, which wasn't changed in 2.6.35 or 2.6.38.

[fewt]

PTS doesn't do anything with measuring wake ups, etc. All that PTS is doing is monitoring the sysfs interface for the power usage, which wasn't changed in 2.6.35 or 2.6.38.

Has the commit you are researching been identified yet?

[not.sure]

Don't think it's with wakeups. I have roughly the same (<~50/s) for 2.6.37 and 2.6.38 and the latter still produces more heat. I suspect it's something much more subtle like RAM and cache refreshs or some chipset stuff.

[adriankx]

The beaty of FREE software

its free so no guarantee.So its like go ahead use it, it will only boil ur hardware and destroy it prematurely. No one cares about power management because in linux world u dont have 1 dictator like in windows world. U actually a few of the red hat novell suse. Users are regarded like dumb enought individuals to make a nice testbed for big bucks companies.Linux target is super computers couse there is where the money is. So linux desktop is still like it was 10 year ago, very much behind windows or MacOS. Sure linux dominates server market and supercomputers. So u get what u pay for, since u pay nothing u kindaa get nothing or close to nothing. Never the less linux is an interesting project to watch dual boot from time to time.

I had sacrificed 1 dell latitude d810 with ati dedicated video card to linux. Yes it got fried in 8 months silly me thinking that always higher temps than windows dont affect hardware. In the end videochipset from all the overheating just gone dead , meaning 10 min of usage until it froze with desktop coruption then 30 bin break to be able to run linux for 10 min more.

In the end i learn my lesson recently i purchased a new laptop AMD based unfortunatly but i bought a notebook cooling pad from day 1. U guessed why because i wanted to use linux somehow. So until those asshats say is time for power optimisation on linux it aint gonna happen. Pretty much like if Bill gates says no than its a no go.

I`m sure many of u will feel ofended, linux is a nice sideproject but nothing more! and used regulary can damage ur hardware prematurely. Hope all of u have quad cores or hexa cores to play. Unfortunatly i have a 1,5 dual core athlon neo x2 thinkpad edge. So i can feel the hit of the new 2.6.38 kernel.

Just for record i used linux first in 2003 and try all there is debian based distros , rpm distros, slack and archlinux. Lately temps are higher even on archlinux, havent tried slackware tho with this latest kernel!

Wish u all the best and hopefully if cry out loud enought the dictators of linux developement and kernel will hear us.

[locovaca]

I had sacrificed 1 dell latitude d810 with ati dedicated video card to linux. Yes it got fried in 8 months silly me thinking that always higher temps than windows dont affect hardware. In the end videochipset from all the overheating just gone dead , meaning 10 min of usage until it froze with desktop coruption then 30 bin break to be able to run linux for 10 min more.

Then you have a problem with Dell, not with Linux. Even if Linux had no power management whatsoever and ran everything full-bore every laptop is supposed to be designed to withstand that load, either by dissipating the heat or shutting down when critical temps were reached.

You could just as easily reach critical temps with Windows by doing video ripping and playback, playing games, browsing a site with Flash, anything that taxes the CPU and GPU for an extended period of time. A cooling system unable to handle that is a Dell issue, not a Linux issue.

Besides, I ran Linux for 5 years on my D800 with a discrete Nvidia Quadro card. I sold it last year with Linux on it, no complaints of it mysteriously dying.

[RealNC]

You could just as easily reach critical temps with Windows by doing video ripping and playback, playing games, browsing a site with Flash, anything that taxes the CPU and GPU for an extended period of time.

There's a *huge* difference between always running hot and only when it's needed. It can mean a difference of double or tripple (or more) the life span of a piece of hardware.

[locovaca]

There's a *huge* difference between always running hot and only when it's needed. It can mean a difference of double or tripple (or more) the life span of a piece of hardware.

Processors have a maximum TDP, as do GPUs. Cooling systems are designed to remove that much heat from the system. If they can't, you WILL reach a critical temperature and the system will shut down.

Please provide a link to your claim of the decreased lifespan. I ran Folding/RC5 for years on all my hardware, laptops included, powered up 24/7. I compiled entire Gentoo installations locally on laptops many times over, back when OO took a full day to emerge. I have not had any failures of chipsets, CPUs, or Video Cards because of it. Dual celerons, dual athlon MPs, dual opeterons, P3-Mobile, P4-Mobile, Pentium M, C2D, none of it failed.