Linux 4.17 Offers Some Promising Power-Savings Improvements

[phoronix]

Phoronix: Linux 4.17 Offers Some Promising Power-Savings Improvements

Of the many improvements to be found in the in-development Linux 4.17 kernel -- nicely summarized in our Linux 4.17 feature overview -- one of the features I've been anxious the most to begin benchmarking has been the reported power management improvements. Here are my initial power/performance tests of Linux 4.17 that for some systems is seeing a measurable drop in power usage, even in some cases under load while without sacrificing the performance.

http://www.phoronix.com/vr.php?view=26206

[Veerappan]

Wow. Looking forward to trying this out on my quad haswell mobile at work (thinkpad T440p).might finally be able to get it down below 10w at idle

[schmidtbag]

Wow that's some pretty good results with practically no regressions. This is some pretty exciting results. A little strange how AMD didn't seem to be affected, but I guess it's not too surprising.

[fasterday]

It looks very promising ! It is good to see some improvements in energy management.

[microcode]

Very cool, and it seems like it's generic (in fact, originally intended for ARM SoCs). A generic core kernel change yielding an almost 20% reduction in total idle system power dissipation on a real laptop is largely unheard of. I wonder how long it'll take for this to trickle down to the aging Android kernel trees (where major SoCs ship today with kernel 4.4 (from January 2016) in the BSP, and preexisting SoCs running Oreo can have a kernel as old as 3.18 [from December 2014]).

[lucasbekker]

Backport please...

[microcode]

Backport please...

This is what poor Greg K.H. is always on about. For some reason, people can't wrap their head around testing and shipping up to date stable and LTS kernels, so they rely on backporting. The problem is, backporting is so labour intensive that very little of it is ever done, in proportion to the huge number of upstream patches which people would like to see on their devices.

With Android and SoC vendors, it has become so severe that Greg says vendors which are trying honestly to fix the upstream gap are five or six years away from achieving that.

[lucasbekker]

This is what poor Greg K.H. is always on about. For some reason, people can't wrap their head around testing and shipping up to date stable and LTS kernels, so they rely on backporting. The problem is, backporting is so labour intensive that very little of it is ever done, in proportion to the huge number of upstream patches which people would like to see on their devices.

With Android and SoC vendors, it has become so severe that Greg says vendors which are trying honestly to fix the upstream gap are five or six years away from achieving that.

That is sad news (for me at least).

Let's just hope that the good people at canonical decide to backport it to their upcoming 18.04 LTS release, providing this significant power saving to people not riding the HWE stack...

[starshipeleven]

Very cool, and it seems like it's generic (in fact, originally intended for ARM SoCs). A generic core kernel change yielding an almost 20% reduction in total idle system power dissipation on a real laptop is largely unheard of.

Also for large multicore servers. 10-15 watts less on load is significant, it gives more headroom for boost states.

[RSpliet]

Michael if I may take the liberty of forwarding, translating and reframing a question asked on Tweakers as a result of your measurement: how much effect does the reworked "idle loop" code have on embedded ARM platforms? Can we somehow quantify what a difference this could make to the battery life of Android phones if back-ported to their kernels?