[Powertop] powertop putting far too much blame on wifi

[Powertop] powertop putting far too much blame on wifi

[Peter F. Patel-Schneider]

[-- Attachment #1: Type: text/plain, Size: 1713 bytes --]

Hi:

I just got a Lenovo Yoga 2 Pro, and I'mrunning Fedora 20 on it.

I'm trying to get power usage down with powertop, but it is producing some 
screwy numbers.   When I run the machine with wifi on, I get idle consumption 
just under 7.2W but almost 5W of that is assigned to the WiFi interface, even 
though there is little traffic. When I turn the wifi off, power consumption 
only drops to just under 7W.  I don't think that there is a phantom WiFi that 
is still consuming power, so it appears to me that powertop is assigning too 
much blame to the WiFi interface.

Is there away of fixing the assignment of blame here?

peter

PS: Here are truncated typical powertop outputs, first with wifi on and then 
with wifi off.

1/
The battery reports a discharge rate of 7.15 W
The estimated remaining time is 3 hours, 39 minutes

Summary: 251.9 wakeups/second, 4.2 GPU ops/seconds, 0.0 VFS ops/sec and 1.6% 
CPU use

Power est. Usage Events/s Category Description
   4.82 W 1.2 pkts/s Device Network interface: wlp1s0 (iwlwifi)
   1.72 W 10.0% Device Display backlight
   131 mW 13.1 ms/s 188.4 Process /usr/lib64/thunderbird/thunderbird
  11.7 mW 398.5 µs/s 19.9 Process [irq/61-iwlwifi]


2/
The battery reports a discharge rate of 6.92 W
The estimated remaining time is 3 hours, 42 minutes

Summary: 453.6 wakeups/second, 23.2 GPU ops/seconds, 0.0 VFS ops/sec and 3.0% 
CPU use

Power est. Usage Events/s Category Description
   1.47 W 10.0% Device Display backlight
   205 mW 14.4 ms/s 197.5 Process /usr/lib64/thunderbird/thunderbird
   184 mW 2.2 ms/s 34.1 Process /usr/bin/xfce4-terminal
   110 mW 2.4 ms/s 164.4 Interrupt PS/2 Touchpad / Keyboard / Mouse

Re: [Powertop] powertop putting far too much blame on wifi

[Ivan Shapovalov]

[-- Attachment #1: Type: text/plain, Size: 2214 bytes --]

On Monday 30 December 2013 at 09:36:08, Peter wrote:
> Hi:
> 
> I just got a Lenovo Yoga 2 Pro, and I'mrunning Fedora 20 on it.
> 
> I'm trying to get power usage down with powertop, but it is producing some
> screwy numbers.   When I run the machine with wifi on, I get idle
> consumption just under 7.2W but almost 5W of that is assigned to the WiFi
> interface, even though there is little traffic. When I turn the wifi off,
> power consumption only drops to just under 7W.  I don't think that there is
> a phantom WiFi that is still consuming power, so it appears to me that
> powertop is assigning too much blame to the WiFi interface.
> 
> Is there away of fixing the assignment of blame here?
> 
> peter
> 
> PS: Here are truncated typical powertop outputs, first with wifi on and then
> with wifi off.
> 
> 1/
> The battery reports a discharge rate of 7.15 W
> The estimated remaining time is 3 hours, 39 minutes
> 
> Summary: 251.9 wakeups/second, 4.2 GPU ops/seconds, 0.0 VFS ops/sec and 1.6%
> CPU use
> 
> Power est. Usage Events/s Category Description
>    4.82 W 1.2 pkts/s Device Network interface: wlp1s0 (iwlwifi)
>    1.72 W 10.0% Device Display backlight
>    131 mW 13.1 ms/s 188.4 Process /usr/lib64/thunderbird/thunderbird
>   11.7 mW 398.5 µs/s 19.9 Process [irq/61-iwlwifi]
> 
> 
> 2/
> The battery reports a discharge rate of 6.92 W
> The estimated remaining time is 3 hours, 42 minutes
> 
> Summary: 453.6 wakeups/second, 23.2 GPU ops/seconds, 0.0 VFS ops/sec and
> 3.0% CPU use
> 
> Power est. Usage Events/s Category Description
>    1.47 W 10.0% Device Display backlight
>    205 mW 14.4 ms/s 197.5 Process /usr/lib64/thunderbird/thunderbird
>    184 mW 2.2 ms/s 34.1 Process /usr/bin/xfce4-terminal
>    110 mW 2.4 ms/s 164.4 Interrupt PS/2 Touchpad / Keyboard / Mouse
> 
> _______________________________________________
> PowerTop mailing list
> PowerTop(a)lists.01.org
> https://lists.01.org/mailman/listinfo/powertop

Yeah. The computation algorithm seems totally off for me.

Currently I get 16.8W for backlight while total is 24W,
and CPU is reported to consume 0W despite ~70% usage.

-- 
Ivan Shapovalov / intelfx /

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 230 bytes --]

Re: [Powertop] powertop putting far too much blame on wifi

[Marc MERLIN]

[-- Attachment #1: Type: text/plain, Size: 2308 bytes --]

On Fri, Jan 03, 2014 at 04:01:27PM +0400, Ivan Shapovalov wrote:
> On Monday 30 December 2013 at 09:36:08, Peter wrote:
> > Hi:
> > 
> > I just got a Lenovo Yoga 2 Pro, and I'mrunning Fedora 20 on it.
> > 
> > I'm trying to get power usage down with powertop, but it is producing some
> > screwy numbers.   When I run the machine with wifi on, I get idle
> > consumption just under 7.2W but almost 5W of that is assigned to the WiFi
> > interface, even though there is little traffic. When I turn the wifi off,
> > power consumption only drops to just under 7W.  I don't think that there is
> > a phantom WiFi that is still consuming power, so it appears to me that
> > powertop is assigning too much blame to the WiFi interface.
> > 
> > Is there away of fixing the assignment of blame here?

Unfortunately, it's not just you. I've reported the same problem
multiple times over the last year, but not heard back from anyone yet
(except you saying you see the same).

I had this on a T530, and now it's even worse on a T540:
The battery reports a discharge rate of 28.5 W
The estimated remaining time is 1 hours, 53 minutes

Summary: 689.5 wakeups/second,  214.0 GPU ops/seconds, 0.0 VFS ops/sec and 10.4% CPU use

Power est.              Usage       Events/s    Category       Description
  15.3 W     95.8%                      Device         Display backlight
  2.07 W     36.0 ms/s     642.4        Process        /usr/bin/enlightenment
  115 mW      4.5 ms/s      62.3        Interrupt      PS/2 Touchpad / Keyboard / Mouse
 86.8 mW      0.9 ms/s      37.0        Process        xfce4-terminal -T window11 --role=window11 --tab
 71.3 mW     35.4 ms/s      36.2        Process        /usr/bin/X :0 vt7 -br -nolisten tcp -auth /var/r
 51.2 mW      0.8 ms/s      27.9        Timer          hrtimer_wakeup

15W for my backlight at mid brightness? I don't think so.

Last time, I was getting unrealistic values for a USB key, then other
things, it's just all over the map :(

Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/                         | PGP 1024R/763BE901

Re: [Powertop] powertop putting far too much blame on wifi

[Andreas Mohr]

[-- Attachment #1: Type: text/plain, Size: 2269 bytes --]

Hi,

On Sun, Jan 05, 2014 at 12:00:02PM -0800, powertop-request(a)lists.01.org wrote:
> From: Marc MERLIN <marc_powertop(a)merlins.org>

> Unfortunately, it's not just you. I've reported the same problem
> multiple times over the last year, but not heard back from anyone yet
> (except you saying you see the same).
> 
> I had this on a T530, and now it's even worse on a T540:
> The battery reports a discharge rate of 28.5 W
> The estimated remaining time is 1 hours, 53 minutes
> 
> Summary: 689.5 wakeups/second,  214.0 GPU ops/seconds, 0.0 VFS ops/sec and 10.4% CPU use
> 
> Power est.              Usage       Events/s    Category       Description
>   15.3 W     95.8%                      Device         Display backlight
>   2.07 W     36.0 ms/s     642.4        Process        /usr/bin/enlightenment
>   115 mW      4.5 ms/s      62.3        Interrupt      PS/2 Touchpad / Keyboard / Mouse
>  86.8 mW      0.9 ms/s      37.0        Process        xfce4-terminal -T window11 --role=window11 --tab
>  71.3 mW     35.4 ms/s      36.2        Process        /usr/bin/X :0 vt7 -br -nolisten tcp -auth /var/r
>  51.2 mW      0.8 ms/s      27.9        Timer          hrtimer_wakeup
> 
> 15W for my backlight at mid brightness? I don't think so.

600+ wakeups/second sounds like the usual intermediate activity
due to touchpad/mouse movement (that figure is exactly what I usually got
on my machine), and if that is the case
(what happens if the human part gets its dirty fingers off the machine? :)
then it's strange to see this otherwise *very high* value (642.4)
attributed to /usr/bin/enlightenment (a very seasoned/stable WM, for
crying out loud) rather than mouse as in the *directly following*
    PS/2 Touchpad / Keyboard / Mouse
line.

Or, in other words, this example hints at a rather mischievous mixup of data,
e.g. an off-by-1 or a post-insert of certain items without accounting for other
columns (within the data model, of course), or perhaps (the column data values
might even support such a suspicion) even completely wrong (re-sorted)
display of data.

Perhaps I should actually test/investigate a recent powertop version... ;)

Andreas Mohr

-- 
GNU/Linux. It's not the software that's free, it's you.

Re: [Powertop] powertop putting far too much blame on wifi

[Ivan Shapovalov]

[-- Attachment #1: Type: text/plain, Size: 2591 bytes --]

On Monday 06 January 2014 at 11:53:15, Andreas wrote:
> Hi,
> 
> On Sun, Jan 05, 2014 at 12:00:02PM -0800, powertop-request(a)lists.01.org 
wrote:
> > From: Marc MERLIN <marc_powertop(a)merlins.org>
> > 
> > Unfortunately, it's not just you. I've reported the same problem
> > multiple times over the last year, but not heard back from anyone yet
> > (except you saying you see the same).
> > 
> > I had this on a T530, and now it's even worse on a T540:
> > The battery reports a discharge rate of 28.5 W
> > The estimated remaining time is 1 hours, 53 minutes
> > 
> > Summary: 689.5 wakeups/second,  214.0 GPU ops/seconds, 0.0 VFS ops/sec and
> > 10.4% CPU use
> > 
> > Power est.              Usage       Events/s    Category       Description
> > 
> >   15.3 W     95.8%                      Device         Display backlight
> >   2.07 W     36.0 ms/s     642.4        Process       
> >   /usr/bin/enlightenment
> >   115 mW      4.5 ms/s      62.3        Interrupt      PS/2 Touchpad /
> >   Keyboard / Mouse>  
> >  86.8 mW      0.9 ms/s      37.0        Process        xfce4-terminal -T
> >  window11 --role=window11 --tab 71.3 mW     35.4 ms/s      36.2       
> >  Process        /usr/bin/X :0 vt7 -br -nolisten tcp -auth /var/r 51.2 mW 
> >      0.8 ms/s      27.9        Timer          hrtimer_wakeup> 
> > 15W for my backlight at mid brightness? I don't think so.
> 
> 600+ wakeups/second sounds like the usual intermediate activity
> due to touchpad/mouse movement (that figure is exactly what I usually got
> on my machine), and if that is the case
> (what happens if the human part gets its dirty fingers off the machine? :)
> then it's strange to see this otherwise *very high* value (642.4)
> attributed to /usr/bin/enlightenment (a very seasoned/stable WM, for
> crying out loud) rather than mouse as in the *directly following*
>     PS/2 Touchpad / Keyboard / Mouse
> line.
> 
> Or, in other words, this example hints at a rather mischievous mixup of
> data, e.g. an off-by-1 or a post-insert of certain items without accounting
> for other columns (within the data model, of course), or perhaps (the
> column data values might even support such a suspicion) even completely
> wrong (re-sorted) display of data.
> 
> Perhaps I should actually test/investigate a recent powertop version... ;)
> 
> Andreas Mohr

IMHO, the usage and event counters seem to be valid. What's wrong is energy 
estimation: it is done entirely in software, using some strange semi-heuristic 
guesswork.

-- 
Ivan Shapovalov / intelfx /

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 230 bytes --]

Re: [Powertop] powertop putting far too much blame on wifi

[Alexandra Yates]

[-- Attachment #1: Type: text/plain, Size: 3001 bytes --]

> On Monday 06 January 2014 at 11:53:15, Andreas wrote:
>> Hi,
>>
>> On Sun, Jan 05, 2014 at 12:00:02PM -0800, powertop-request(a)lists.01.org
> wrote:
>> > From: Marc MERLIN <marc_powertop(a)merlins.org>
>> >
>> > Unfortunately, it's not just you. I've reported the same problem
>> > multiple times over the last year, but not heard back from anyone yet
>> > (except you saying you see the same).
>> >
>> > I had this on a T530, and now it's even worse on a T540:
>> > The battery reports a discharge rate of 28.5 W
>> > The estimated remaining time is 1 hours, 53 minutes
>> >
>> > Summary: 689.5 wakeups/second,  214.0 GPU ops/seconds, 0.0 VFS ops/sec
>> and
>> > 10.4% CPU use
>> >
>> > Power est.              Usage       Events/s    Category
>> Description
>> >
>> >   15.3 W     95.8%                      Device         Display
>> backlight
>> >   2.07 W     36.0 ms/s     642.4        Process
>> >   /usr/bin/enlightenment
>> >   115 mW      4.5 ms/s      62.3        Interrupt      PS/2 Touchpad /
>> >   Keyboard / Mouse>
>> >  86.8 mW      0.9 ms/s      37.0        Process        xfce4-terminal
>> -T
>> >  window11 --role=window11 --tab 71.3 mW     35.4 ms/s      36.2
>> >  Process        /usr/bin/X :0 vt7 -br -nolisten tcp -auth /var/r 51.2
>> mW
>> >      0.8 ms/s      27.9        Timer          hrtimer_wakeup>
>> > 15W for my backlight at mid brightness? I don't think so.
>>
>> 600+ wakeups/second sounds like the usual intermediate activity
>> due to touchpad/mouse movement (that figure is exactly what I usually
>> got
>> on my machine), and if that is the case
>> (what happens if the human part gets its dirty fingers off the machine?
>> :)
>> then it's strange to see this otherwise *very high* value (642.4)
>> attributed to /usr/bin/enlightenment (a very seasoned/stable WM, for
>> crying out loud) rather than mouse as in the *directly following*
>>     PS/2 Touchpad / Keyboard / Mouse
>> line.
>>
>> Or, in other words, this example hints at a rather mischievous mixup of
>> data, e.g. an off-by-1 or a post-insert of certain items without
>> accounting
>> for other columns (within the data model, of course), or perhaps (the
>> column data values might even support such a suspicion) even completely
>> wrong (re-sorted) display of data.
>>
>> Perhaps I should actually test/investigate a recent powertop version...
>> ;)
>>
>> Andreas Mohr
>
> IMHO, the usage and event counters seem to be valid. What's wrong is
> energy
> estimation: it is done entirely in software, using some strange
> semi-heuristic
> guesswork.
>
> --
> Ivan Shapovalov / intelfx /_______________________________________________
> PowerTop mailing list
> PowerTop(a)lists.01.org
> https://lists.01.org/mailman/listinfo/powertop
>

Try disabling the device and rerunning PowerTOP see if there is a
difference there. You could send the debug output to find out what are
your wifi parameters.

Thank you,
Alexandra.

Re: [Powertop] powertop putting far too much blame on wifi

[Marc MERLIN]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

On Tue, Jan 07, 2014 at 03:30:38PM -0800, Alexandra Yates wrote:
> > IMHO, the usage and event counters seem to be valid. What's wrong is
> > energy
> > estimation: it is done entirely in software, using some strange
> > semi-heuristic
> > guesswork.

For me, the most likely incorrect usage report just shifts to another
line if it's a device I can remove (although most often it's integrated
in the laptop, so I can't remove it).

Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/                         | PGP 1024R/763BE901