[Intel-gfx] [PATCH 2/3] drm/i915: Boost RPS frequency for CPU stalls

[Jesse Barnes]

On Wed, 2 Oct 2013 15:18:20 -0700
Jesse Barnes <jbarnes at virtuousgeek.org> wrote:

> On Wed, 2 Oct 2013 22:57:41 +0100
> Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > No, that is a factor I had not considered. I had been concerned about
> > monitoring latency whilst minimising frequency and using the power gauge
> > as the ultimate measure of success. One thing that is apparent with ivb,
> > is that the power gauge (at least) is dependent upon workload. That is
> > you can set high frequencies, but if the ring is idle (but not in rc6)
> > then it consumes relatively little power. (That still may be 1W on the
> > big GPUs, but it does mean that at low frequencies there is not a lot of
> > difference between rc6 on/off, a few 10s of mW, and we are concerned
> > with a window of about 10ms.) I suspect the difference is likely to be
> > in the noise for an idle desktop workload and would only really show
> > itself in a synthetic benchmark.
> 
> Hm are you sure about that?  It doesn't match the last measurements I
> had from a long time ago... until RC6 kicked in the GPU burned quite a
> bit (several Watts).
> 
> If you're correct though, then I agree, this is quibbling over crumbs,
> so not worth modifying.

Oh and the other thing that comes to mind is platform power.  Even if
RC6 vs no RC6 at low freq is minimally different, entering RC6 allows
the platform to go into a deep sleep state if the CPUs are also idle.
So race to idle as efficiently as possible is probably still the right
answer.

Jesse