[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Power-Management

Hi Marc.

Marc Haber - 26.04.18, 11:00:
> On Thu, Apr 26, 2018 at 09:35:45AM +0100, Chris Lamb wrote:
> > > I was just Wondering if you have any considerations for issuing an
> > > update or designing lets say Debian 10 with Power Management for
> > > laptops in mind.
> > 
> > Such modifications are typically (and best) made in a distribution-
> > agnostic manner, rather than being, say, Debian-specific. :)
> > 
> > However, may I take this opportunity to promote Jonathan Carter aka
> > highvoltage's "Debian Package of the Day" video series? As it
> > 
> > happens, yesterday's package was "powertop":
> >   https://www.youtube.com/watch?v=p3M_g3C15R4
> Power Management is a general big issue in Linux on Laptops. I have
> also made the experience that Windows achives vastly better battery
> run times than any Linux I have ever encountered, even if powertop
> doesn't have anything to complain.
> Alas, I don't have the slightest idea why, maybe somebody can shed a
> light on that.

I think I can only offer what may be a part of the answer to this 

Part of the reason is SATA device link power management (LPM). In 
Windows as far as I concluded about what I read about Linux patches it 
has been enabled for a long time. However as kernel developers enabled 
it, some devices failed badly and AFAIR there even have been corrupted 
data. So kernel developers backed out. Only recently, the implemented 
another link power management mode, similar to what Windows does, and 
this appears to work. Except where it didn´t, like for example my 
Crucial m500 mSATA SSD in this ThinkPad T520, which meanwhile is 
blacklisted. If I remember LPM can make at a difference of a few watts, 
which is quite something. I don´t have links at hand, but I bet articles 
or mailing list posts about that are easy enough to find.

From what I read over the time I concluded that this could be part of a 
more generic reason: Power saving mechanism that work on most devices, 
do not work on all devices. Or need to be implemented in a way that is 
difficult to get right. And I think in part this is just due to badly 
implemented or even buggy firmware in hardware. And in part also due to 
firmwares being closed source blobs, which only the hardware 
manufacturer probably knowing what it really does. Like with 
optimization for SSDs, in part it is all just guess work.

I think what would be beneficial here, if some vendor starts to build 
laptops *from the ground up* for the usage with free software operating 
systems, by opening up all firmware, documenting their interfaces, 
especially for power management. But this would have to be for all 
devices in a laptop, and while there are open channel SSDs for special 
server / data center workloads which just expose their flash chips and 
let Linux do all the rest, there is still so much closed firmware just 
about everywhere. I recently researched how to obtain a laptop with as 
much free software firmware as I can get, put apart from the offers by 
puri.sm, it is a lot of manual work involved or paying someone to do 
that work.

Of cause all that said, I still believe that it is possible to improve 
things on the Linux side. But this does not only need kernel work, it 
also needs user space work. Desktop environments with many features like 
Plasma or GNOME IMO can still be optimized a lot to wake up less, do 
less I/O (especially KDEPIM + Akonadi), and use power saving features 
more aggressively. People with tiling window managers and minimal, 
mostly console tool based desktops may have an advantage here.


Reply to: