Final set of results from the Power Management Blueprint

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Final set of results from the Power Management Blueprint

Colin Ian King-2
Hi,

Now beta2 is nearly here I ran some (very simple) tests last night and
this morning on Unity 2D vs 3D to see how they compare in terms of power
consumption on a HP Mini 210 (Atom N450) and a Lenovo X220i
(Sandybridge, i3).

Data:
     
http://zinc.canonical.com/~cking/power-benchmarking/unity-beta2-results/unity-2d-vs-3d.ods

Results:
     
http://zinc.canonical.com/~cking/power-benchmarking/unity-beta2-results/unity-beta2-2d-vs-3d-results.txt

Quick Summary:

For 3D animation and busy rendering (which we suspect this keeps the
compositor busy) Unity 3D consumes more power than 2D.  This is more
apparent on newer Intel Sandybridge chipsets because 2D seems to keep
the GPU powered down which is known to save 25-40% of total system power
due to the i915 RC6 power saving mode.

For some operations, such as desktop zooming, 2D is more expensive
because we don't have GPU assist to scale images.

Finally, for idle systems, 2D and 3D are comparable in power consumption.

..but see the unity-beta2-2d-vs-3d-results.txt for more details.  Oh,
and the sigma (std.deviation) is rendered poorly on web-browsers.

Colin

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

Re: [Canonical-tech] Final set of results from the Power Management Blueprint

Mark Shuttleworth-3
On 28/03/12 14:46, Colin Ian King wrote:
> Now beta2 is nearly here I ran some (very simple) tests last night and
> this morning on Unity 2D vs 3D to see how they compare in terms of
> power consumption on a HP Mini 210 (Atom N450) and a Lenovo X220i
> (Sandybridge, i3).

Awesome work, Colin. Thank you! Both for the reporting and the
improvements in 12.04.

Should we be worried about the 100mA idle delta between 2D and 3D on the
Lenovo?

Mark

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

Re: [Canonical-tech] Final set of results from the Power Management Blueprint

Colin Ian King-2
On 28/03/12 17:09, Mark Shuttleworth wrote:

> On 28/03/12 14:46, Colin Ian King wrote:
>> Now beta2 is nearly here I ran some (very simple) tests last night and
>> this morning on Unity 2D vs 3D to see how they compare in terms of
>> power consumption on a HP Mini 210 (Atom N450) and a Lenovo X220i
>> (Sandybridge, i3).
>
> Awesome work, Colin. Thank you! Both for the reporting and the
> improvements in 12.04.
>
> Should we be worried about the 100mA idle delta between 2D and 3D on the
> Lenovo?

It is of interest to find out why this is, so I will see if I can find
out what is the root cause, but it could be anywhere in the layers of
sub-systems.

Colin

>
> Mark


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

Re: [Canonical-tech] Final set of results from the Power Management Blueprint

Colin Ian King-2
In reply to this post by Mark Shuttleworth-3
On 28/03/12 17:09, Mark Shuttleworth wrote:

> On 28/03/12 14:46, Colin Ian King wrote:
>> Now beta2 is nearly here I ran some (very simple) tests last night and
>> this morning on Unity 2D vs 3D to see how they compare in terms of
>> power consumption on a HP Mini 210 (Atom N450) and a Lenovo X220i
>> (Sandybridge, i3).
>
> Awesome work, Colin. Thank you! Both for the reporting and the
> improvements in 12.04.
>
> Should we be worried about the 100mA idle delta between 2D and 3D on the
> Lenovo?

I've dug into this by comparing wakeup events and process activity over
100 seconds of idle time for 2D vs 3D.  The only easily noticeable
difference between 2D and 3D in this respect is that 3D has compiz
running and it wakes up ~2-3 times a second on average. The CPU
utilisation for this is very small though, ~0.03%, so I am supposing
there may be some GPU activity going on too which may consume some
power. I'm not sure how easy it would be to verify this.

Also, Xorg in 3D seems to be using a very small amount more CPU in 3D
compared to 2D so that may account for a small part of the 100mA.

Apart from that, I'm finding it hard to account as there is quite a bit
of 'noise' in the data I've captured which affects the final analysis.

Colin


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team