i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

Hi,

Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on
Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver
loaded; brightness set to 4/100; X running; no desktop environment, except
Awesome), from 6.5W to about 10.5W, as reported by powertop.

In the stable tree, it bisects to:
        commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
        Author: Chris Wilson <chris@chris-wilson.co.uk>
        Date:   Fri Jun 6 10:37:11 2014 +0100

            drm/i915: Disable FBC by default also on Haswell and later

            commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.

            It causes black screen on bootup and is approximately 100x
slower than
            running with FBC disabled, so the GPU runs at a high
frequency for much
            longer - completely contrary to the power saving claims.
It also still
            has mutex deadlocks in multi-head scenarios, which can lead to a
            system/X lockup. These bugs were known before FBC was
enabled by default
            on Haswell and still have not been fixed.

The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).

With a 75Wh battery, that's a significant loss in battery life in normal use.

I'll be happy to help test any potential fix.

Thanks,
Eric

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Jani Nikula]

On Fri, 22 Aug 2014, Eric Rannaud <eric.rannaud@gmail.com> wrote:
> Hi,
>
> Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on
> Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver
> loaded; brightness set to 4/100; X running; no desktop environment, except
> Awesome), from 6.5W to about 10.5W, as reported by powertop.
>
> In the stable tree, it bisects to:
>         commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
>         Author: Chris Wilson <chris@chris-wilson.co.uk>
>         Date:   Fri Jun 6 10:37:11 2014 +0100
>
>             drm/i915: Disable FBC by default also on Haswell and later
>
>             commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.
>
>             It causes black screen on bootup and is approximately 100x
> slower than
>             running with FBC disabled, so the GPU runs at a high
> frequency for much
>             longer - completely contrary to the power saving claims.
> It also still
>             has mutex deadlocks in multi-head scenarios, which can lead to a
>             system/X lockup. These bugs were known before FBC was
> enabled by default
>             on Haswell and still have not been fixed.
>
> The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).
>
> With a 75Wh battery, that's a significant loss in battery life in normal use.
>
> I'll be happy to help test any potential fix.

The earlier regression trumps, and in this case it was enabling FBC by
default on Haswell. Sorry.

You can enable FBC with i915.enable_fbc=1 module parameter, but all bets
are off. See the commit message you quoted above. I don't recommend.

BR,
Jani.


-- 
Jani Nikula, Intel Open Source Technology Center

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Jani Nikula]

[just moving from lkml to intel-gfx for a better fitting audience]

On Mon, 25 Aug 2014, Jani Nikula <jani.nikula@intel.com> wrote:
> On Fri, 22 Aug 2014, Eric Rannaud <eric.rannaud@gmail.com> wrote:
>> Hi,
>>
>> Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on
>> Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver
>> loaded; brightness set to 4/100; X running; no desktop environment, except
>> Awesome), from 6.5W to about 10.5W, as reported by powertop.
>>
>> In the stable tree, it bisects to:
>>         commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
>>         Author: Chris Wilson <chris@chris-wilson.co.uk>
>>         Date:   Fri Jun 6 10:37:11 2014 +0100
>>
>>             drm/i915: Disable FBC by default also on Haswell and later
>>
>>             commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.
>>
>>             It causes black screen on bootup and is approximately 100x
>> slower than
>>             running with FBC disabled, so the GPU runs at a high
>> frequency for much
>>             longer - completely contrary to the power saving claims.
>> It also still
>>             has mutex deadlocks in multi-head scenarios, which can lead to a
>>             system/X lockup. These bugs were known before FBC was
>> enabled by default
>>             on Haswell and still have not been fixed.
>>
>> The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).
>>
>> With a 75Wh battery, that's a significant loss in battery life in normal use.
>>
>> I'll be happy to help test any potential fix.
>
> The earlier regression trumps, and in this case it was enabling FBC by
> default on Haswell. Sorry.
>
> You can enable FBC with i915.enable_fbc=1 module parameter, but all bets
> are off. See the commit message you quoted above. I don't recommend.
>
> BR,
> Jani.
>
>
> -- 
> Jani Nikula, Intel Open Source Technology Center

-- 
Jani Nikula, Intel Open Source Technology Center

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

Hi Jani,

Is there a way to restore the prior lower power consumption when
idling? As I understand it, FBC should not have a direct effect on
power consumption when idle, only when the FB is actively refreshed.
Is it understood why no FBC would have such a dramatic impact (+4W) on
a system sitting idle?

I'm aware of i915.i915_enable_rc6=1 and i915.lvds_downclock=1, which I
haven't yet tried. Is there anything else I should also try?

Thanks,
Eric

On Mon, Aug 25, 2014 at 3:22 AM, Jani Nikula <jani.nikula@intel.com> wrote:
>
> [just moving from lkml to intel-gfx for a better fitting audience]
>
> On Mon, 25 Aug 2014, Jani Nikula <jani.nikula@intel.com> wrote:
>> On Fri, 22 Aug 2014, Eric Rannaud <eric.rannaud@gmail.com> wrote:
>>> Hi,
>>>
>>> Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on
>>> Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver
>>> loaded; brightness set to 4/100; X running; no desktop environment, except
>>> Awesome), from 6.5W to about 10.5W, as reported by powertop.
>>>
>>> In the stable tree, it bisects to:
>>>         commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
>>>         Author: Chris Wilson <chris@chris-wilson.co.uk>
>>>         Date:   Fri Jun 6 10:37:11 2014 +0100
>>>
>>>             drm/i915: Disable FBC by default also on Haswell and later
>>>
>>>             commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.
>>>
>>>             It causes black screen on bootup and is approximately 100x
>>> slower than
>>>             running with FBC disabled, so the GPU runs at a high
>>> frequency for much
>>>             longer - completely contrary to the power saving claims.
>>> It also still
>>>             has mutex deadlocks in multi-head scenarios, which can lead to a
>>>             system/X lockup. These bugs were known before FBC was
>>> enabled by default
>>>             on Haswell and still have not been fixed.
>>>
>>> The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).
>>>
>>> With a 75Wh battery, that's a significant loss in battery life in normal use.
>>>
>>> I'll be happy to help test any potential fix.
>>
>> The earlier regression trumps, and in this case it was enabling FBC by
>> default on Haswell. Sorry.
>>
>> You can enable FBC with i915.enable_fbc=1 module parameter, but all bets
>> are off. See the commit message you quoted above. I don't recommend.
>>
>> BR,
>> Jani.
>>
>>
>> --
>> Jani Nikula, Intel Open Source Technology Center
>
> --
> Jani Nikula, Intel Open Source Technology Center

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Ville Syrjälä]

On Mon, Aug 25, 2014 at 05:19:49AM -0700, Eric Rannaud wrote:
> Hi Jani,
> 
> Is there a way to restore the prior lower power consumption when
> idling? As I understand it, FBC should not have a direct effect on
> power consumption when idle, only when the FB is actively refreshed.
> Is it understood why no FBC would have such a dramatic impact (+4W) on
> a system sitting idle?

FBC works best when the screen contents don't change. The more activity
on the screen the less effective FBC becomes. 4W sounds way too much for
FBC however. 0.4W is closer to what one might expect from FBC based on my
observations. 4W sounds more like the difference between min vs. max
display brightness to me.

> 
> I'm aware of i915.i915_enable_rc6=1 and i915.lvds_downclock=1, which I
> haven't yet tried. Is there anything else I should also try?
> 
> Thanks,
> Eric
> 
> On Mon, Aug 25, 2014 at 3:22 AM, Jani Nikula <jani.nikula@intel.com> wrote:
> >
> > [just moving from lkml to intel-gfx for a better fitting audience]
> >
> > On Mon, 25 Aug 2014, Jani Nikula <jani.nikula@intel.com> wrote:
> >> On Fri, 22 Aug 2014, Eric Rannaud <eric.rannaud@gmail.com> wrote:
> >>> Hi,
> >>>
> >>> Between 3.15.4 and 3.15.8, there was an increase in idle power consumption on
> >>> Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi driver
> >>> loaded; brightness set to 4/100; X running; no desktop environment, except
> >>> Awesome), from 6.5W to about 10.5W, as reported by powertop.
> >>>
> >>> In the stable tree, it bisects to:
> >>>         commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
> >>>         Author: Chris Wilson <chris@chris-wilson.co.uk>
> >>>         Date:   Fri Jun 6 10:37:11 2014 +0100
> >>>
> >>>             drm/i915: Disable FBC by default also on Haswell and later
> >>>
> >>>             commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.
> >>>
> >>>             It causes black screen on bootup and is approximately 100x
> >>> slower than
> >>>             running with FBC disabled, so the GPU runs at a high
> >>> frequency for much
> >>>             longer - completely contrary to the power saving claims.
> >>> It also still
> >>>             has mutex deadlocks in multi-head scenarios, which can lead to a
> >>>             system/X lockup. These bugs were known before FBC was
> >>> enabled by default
> >>>             on Haswell and still have not been fixed.
> >>>
> >>> The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).
> >>>
> >>> With a 75Wh battery, that's a significant loss in battery life in normal use.
> >>>
> >>> I'll be happy to help test any potential fix.
> >>
> >> The earlier regression trumps, and in this case it was enabling FBC by
> >> default on Haswell. Sorry.
> >>
> >> You can enable FBC with i915.enable_fbc=1 module parameter, but all bets
> >> are off. See the commit message you quoted above. I don't recommend.
> >>
> >> BR,
> >> Jani.
> >>
> >>
> >> --
> >> Jani Nikula, Intel Open Source Technology Center
> >
> > --
> > Jani Nikula, Intel Open Source Technology Center
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Ville Syrjälä
Intel OTC

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

On Tue, Aug 26, 2014 at 4:06 AM, Ville Syrjälä
<ville.syrjala@linux.intel.com> wrote:
> FBC works best when the screen contents don't change. The more activity
> on the screen the less effective FBC becomes. 4W sounds way too much for
> FBC however. 0.4W is closer to what one might expect from FBC based on my
> observations. 4W sounds more like the difference between min vs. max
> display brightness to me.

That's why I'm here, because it is such a dramatic change. All
measurements were taken at constant brightness (4/100) and the
difference is indeed a full 4W, as reported by the battery (and seen
in: powertop(1),
/sys/bus/acpi/drivers/battery/PNP0C0A\:00/power_supply/BAT0/power_now,
and a noticeably faster drop in battery charge over time).

Could the FBC commit somehow also disable RC6, or similar?
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

On Tue, Aug 26, 2014 at 6:38 AM, Eric Rannaud <eric.rannaud@gmail.com> wrote:
> Could the FBC commit somehow also disable RC6, or similar?

I just tried these kernel arguments with no significant effect on
power consumption (i.e. still around 11W in idle, instead of under
7W):
        i915.enable_rc6=1 i915.lvds_downclock=1 pcie_aspm=force

Forcing FBC with i915.enable_fbc=1 brings the idle power consumption
back to under 7W, however.
This is all on 3.15.4-ARCH-00041-gf4db98240ac2.

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Daniel Vetter]

On Tue, Aug 26, 2014 at 4:57 PM, Eric Rannaud <eric.rannaud@gmail.com> wrote:
> On Tue, Aug 26, 2014 at 6:38 AM, Eric Rannaud <eric.rannaud@gmail.com> wrote:
>> Could the FBC commit somehow also disable RC6, or similar?
>
> I just tried these kernel arguments with no significant effect on
> power consumption (i.e. still around 11W in idle, instead of under
> 7W):
>         i915.enable_rc6=1 i915.lvds_downclock=1 pcie_aspm=force
>
> Forcing FBC with i915.enable_fbc=1 brings the idle power consumption
> back to under 7W, however.
> This is all on 3.15.4-ARCH-00041-gf4db98240ac2.

Any significant changes in package C state as reported in powertop?
Indeed fairly impressive how much fbc saves here ...
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

On Tue, Aug 26, 2014 at 1:59 PM, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
>> Forcing FBC with i915.enable_fbc=1 brings the idle power consumption
>> back to under 7W, however.
>> This is all on 3.15.4-ARCH-00041-gf4db98240ac2.
>
> Any significant changes in package C state as reported in powertop?
> Indeed fairly impressive how much fbc saves here ...

Not that I can tell.
    Powertop report with FBC: http://pastebin.com/5qfJKpTQ
    Without FBC: http://pastebin.com/NaYkR4n0

Some highly uneducated guesses on what could explain a +4W jump with no FBC:

#1- The higher DRAM and bus duty cycle during scanout is enough to
prevent some DRAM subsystems from sleeping, by crossing some tight
threshold (maybe Apple has FBC enabled, so parameters somewhere in
firmware are tuned for the lower level of background activity they
expect with FBC on?). Not much we can do about that, unless such
parameters can be tweaked by us.

#2- Without FBC, the FB doesn't fit in L3 (i7-4750HQ has 6MB,
2880x1800 compressed at least 1:4 fits), keeping the DRAM awake more.

#3- Disabling FBC somehow affects the layout of the framebuffer in
DRAM, keeping more of the DRAM active and awake during scanout.
Different tiling, swizzling, etc. parameters? Is it worth looking at
the code for that kind of thing? To be clear, I'm (blindly) suggesting
that it might be possible to increase the locality of the framebuffer
in physical DRAM, even without compression enabled.

Actually, with an image of white-noise displayed fullscreen while
powertop takes a 20 second measurement, the idle power consumption
shoots up to 11W, with FBC enabled. This experiment actually
invalidates my guess #3. The white noise image will not compress much
at all, while my typical test screen is a couple of static
black-and-white terminal windows (filling up the screen), which will
compress well.

So it would appear 4W is the actual cost of having a full-size 20MB
framebuffer on this system.

Anything an outsider can do to help getting FBC enabled by default again?

Thanks,
Eric

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Ville Syrjälä]

On Tue, Aug 26, 2014 at 04:00:51PM -0700, Eric Rannaud wrote:
> On Tue, Aug 26, 2014 at 1:59 PM, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
> >> Forcing FBC with i915.enable_fbc=1 brings the idle power consumption
> >> back to under 7W, however.
> >> This is all on 3.15.4-ARCH-00041-gf4db98240ac2.
> >
> > Any significant changes in package C state as reported in powertop?
> > Indeed fairly impressive how much fbc saves here ...
> 
> Not that I can tell.
>     Powertop report with FBC: http://pastebin.com/5qfJKpTQ
>     Without FBC: http://pastebin.com/NaYkR4n0

Seems to have gotten messed up somehow. I can't see the package c-state
info there at all for some reason. Also core c-states go only up to c7s.
On hsw I would have expected to see deeper c-states, but maybe it's not
a ult machine or something. My hsw-ult goes up to c10.

> Some highly uneducated guesses on what could explain a +4W jump with no FBC:
> 
> #1- The higher DRAM and bus duty cycle during scanout is enough to
> prevent some DRAM subsystems from sleeping, by crossing some tight
> threshold (maybe Apple has FBC enabled, so parameters somewhere in
> firmware are tuned for the lower level of background activity they
> expect with FBC on?). Not much we can do about that, unless such
> parameters can be tweaked by us.
> 
> #2- Without FBC, the FB doesn't fit in L3 (i7-4750HQ has 6MB,
> 2880x1800 compressed at least 1:4 fits), keeping the DRAM awake more.
> 
> #3- Disabling FBC somehow affects the layout of the framebuffer in
> DRAM, keeping more of the DRAM active and awake during scanout.
> Different tiling, swizzling, etc. parameters? Is it worth looking at
> the code for that kind of thing? To be clear, I'm (blindly) suggesting
> that it might be possible to increase the locality of the framebuffer
> in physical DRAM, even without compression enabled.
> 
> Actually, with an image of white-noise displayed fullscreen while
> powertop takes a 20 second measurement, the idle power consumption
> shoots up to 11W, with FBC enabled. This experiment actually
> invalidates my guess #3. The white noise image will not compress much
> at all, while my typical test screen is a couple of static
> black-and-white terminal windows (filling up the screen), which will
> compress well.
> 
> So it would appear 4W is the actual cost of having a full-size 20MB
> framebuffer on this system.

I guess the monster resolution just really hurts w/o fbc. My hsw has
1920x1080 panel which, assuming the same refresh rate, means your
display refresh requires 2.5x the bandwidth mine does. Sadly my hsw
seems to have a a dead battery so i can't check the power consumption
figures. pc7 residency definitely drops quite a bit on that machine if
I disable fbc. I don't remember if I ever measured the effect of fbc
on that thing before the battery died, but on my ivb it's definitely
somewhere in the .3W ballpark, though that machine has an even lower
resolution than the hsw.

> 
> Anything an outsider can do to help getting FBC enabled by default again?

Either someone needs to review my patches or fix the fbc code in
some other way. Even if someone wants to do the frontbuffer tracking
some other way, there was plenty of stuff in my patches that can
still be applied (there were even simple bugfixes included iirc).

-- 
Ville Syrjälä
Intel OTC

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

On Wed, Aug 27, 2014 at 2:17 AM, Ville Syrjälä
<ville.syrjala@linux.intel.com> wrote:
>> Not that I can tell.
>>     Powertop report with FBC: http://pastebin.com/5qfJKpTQ
>>     Without FBC: http://pastebin.com/NaYkR4n0
>
> Seems to have gotten messed up somehow. I can't see the package c-state
> info there at all for some reason. Also core c-states go only up to c7s.
> On hsw I would have expected to see deeper c-states, but maybe it's not
> a ult machine or something. My hsw-ult goes up to c10.

This part (i7-4750HQ) is not ULT.
I don't see any more information in the powertop ncurses UI that is in
the reports on pastebin. To be clear, you're not looking for this?

   Package     0
   Powered On 0.0%
   C1E-HSW    0.1%
   C3-HSW     0.0%
   RC6pp      0.0%
   C7s-HSW    93.9%

   GPU 0
   Powered On 0.3%
   RC6        99.7%
   RC6p       0.0%
   RC6pp      0.0%

>> Anything an outsider can do to help getting FBC enabled by default again?
>
> Either someone needs to review my patches or fix the fbc code in
> some other way. Even if someone wants to do the frontbuffer tracking
> some other way, there was plenty of stuff in my patches that can
> still be applied (there were even simple bugfixes included iirc).

Do you have a link to them, or a branch somewhere?

I also tried enable_psr=1 (on Linus' HEAD), but it has no significant
effect on power consumption on this machine.

Thanks,
Eric
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Sean V Kelley]

[-- Attachment #1.1: Type: text/plain, Size: 2424 bytes --]

On Mon, Aug 25, 2014 at 3:22 AM, Jani Nikula <jani.nikula@intel.com> wrote:

>
> [just moving from lkml to intel-gfx for a better fitting audience]
>
> On Mon, 25 Aug 2014, Jani Nikula <jani.nikula@intel.com> wrote:
> > On Fri, 22 Aug 2014, Eric Rannaud <eric.rannaud@gmail.com> wrote:
> >> Hi,
> >>
> >> Between 3.15.4 and 3.15.8, there was an increase in idle power
> consumption on
> >> Apple Macbook Pro 15 (late 2013) on a freshly booted system (no wifi
> driver
> >> loaded; brightness set to 4/100; X running; no desktop environment,
> except
> >> Awesome), from 6.5W to about 10.5W, as reported by powertop.
> >>
> >> In the stable tree, it bisects to:
> >>         commit f4db98240ac2c6d9d2118c6f82d483ff5293f1ed
> >>         Author: Chris Wilson <chris@chris-wilson.co.uk>
> >>         Date:   Fri Jun 6 10:37:11 2014 +0100
> >>
> >>             drm/i915: Disable FBC by default also on Haswell and later
> >>
> >>             commit 0368920e51ae0cded0eb518c340a4dd17764d461 upstream.
> >>
> >>             It causes black screen on bootup and is approximately 100x
> >> slower than
> >>             running with FBC disabled, so the GPU runs at a high
> >> frequency for much
> >>             longer - completely contrary to the power saving claims.
> >> It also still
> >>             has mutex deadlocks in multi-head scenarios, which can lead
> to a
> >>             system/X lockup. These bugs were known before FBC was
> >> enabled by default
> >>             on Haswell and still have not been fixed.
> >>
> >> The issue is still present in Linus' tree (v3.17-rc1-22-g480cadc2b7e0).
> >>
> >> With a 75Wh battery, that's a significant loss in battery life in
> normal use.
> >>
> >> I'll be happy to help test any potential fix.
> >
> > The earlier regression trumps, and in this case it was enabling FBC by
> > default on Haswell. Sorry.
> >
> > You can enable FBC with i915.enable_fbc=1 module parameter, but all bets
> > are off. See the commit message you quoted above. I don't recommend.
> >
>


For what it's worth, I have a Mid-2014, Macbook Pro Retina (13inch
display), running Archlinux with 3.16.

Definitely, enable_fbc is a win for me and I do manually enable it.  But I
am still seeing what I believe to be
a regression overall of about +4W even with fbc enabled.  Still digging for
clues.

Sean





-- 
Sean V. Kelley <sean.v.kelley@intel.com>
Open Source Technology Center / SSG
Intel Corp.

[-- Attachment #1.2: Type: text/html, Size: 3610 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Eric Rannaud]

On Thu, Aug 28, 2014 at 9:41 AM, Sean V Kelley <sean.v.kelley@intel.com> wrote:
> For what it's worth, I have a Mid-2014, Macbook Pro Retina (13inch display),
> running Archlinux with 3.16.
>
> Definitely, enable_fbc is a win for me and I do manually enable it.  But I
> am still seeing what I believe to be
> a regression overall of about +4W even with fbc enabled.  Still digging for
> clues.

To be clear, because I don't think I listed that case, with
3.17.0-rc2-ARCH-00040-gff0c57ac7043 and i915.enable_fbc=1 the idle
power consumption goes back down to under 7W, which is about back to
normal.

There does seem to be a roughly +500mW regression since 3.13, but it's
a little hard to tell because of the sampling noise in the power
information as reported by powertop.

Is there a systematic, large-scale effort to obtain baseline power
data over time on various configs?

Re: i915: Regression: +4W in idle power use on Macbook Pro 15 (late 2013)

[Lu, Ran]

Hi,

On Tuesday 26 August 2014 16:00:51, Eric Rannaud wrote：
> On Tue, Aug 26, 2014 at 1:59 PM, Daniel Vetter <daniel.vetter@ffwll.ch> 
wrote:
> >> Forcing FBC with i915.enable_fbc=1 brings the idle power consumption
> >> back to under 7W, however.
> >> This is all on 3.15.4-ARCH-00041-gf4db98240ac2.
> > 
> > Any significant changes in package C state as reported in powertop?
> > Indeed fairly impressive how much fbc saves here ...
> 
> Not that I can tell.
>     Powertop report with FBC: http://pastebin.com/5qfJKpTQ
>     Without FBC: http://pastebin.com/NaYkR4n0
> 
> Some highly uneducated guesses on what could explain a +4W jump with no FBC:
> 
> #1- The higher DRAM and bus duty cycle during scanout is enough to
> prevent some DRAM subsystems from sleeping, by crossing some tight
> threshold (maybe Apple has FBC enabled, so parameters somewhere in
> firmware are tuned for the lower level of background activity they
> expect with FBC on?). Not much we can do about that, unless such
> parameters can be tweaked by us.
> 
> #2- Without FBC, the FB doesn't fit in L3 (i7-4750HQ has 6MB,
> 2880x1800 compressed at least 1:4 fits), keeping the DRAM awake more.
> 
> #3- Disabling FBC somehow affects the layout of the framebuffer in
> DRAM, keeping more of the DRAM active and awake during scanout.
> Different tiling, swizzling, etc. parameters? Is it worth looking at
> the code for that kind of thing? To be clear, I'm (blindly) suggesting
> that it might be possible to increase the locality of the framebuffer
> in physical DRAM, even without compression enabled.

I notices similar behavior. From the output of turborstat, with FBC MBP can 
reach PC6 more than 90% of the time when idling, after FBC was disable in 3.14 
or 3.15 the chip stays in PC2 and the estimated GPU power consumption is a few 
watts higher.

-- 
Best Regards,
LR
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx