Power Management Test day (2011-09-29) Stats

[Adam Williamson]

On Thu, 2011-10-06 at 09:56 -0500, Michael Cronenworth wrote:
> Jaroslav Skarvada wrote:
> > I cannot reproduce these numbers on our testing machine - in [1] power
> > consumption in active idle is +- measurement error, in other tests
> > it is mostly higher power consumption, but also higher performance.
> > Similar for idle graph in [2]. I will try to get one of the mentioned
> > machines and recheck. More detailed comparison of power consumption of
> > various Fedoras is on my todo list
> 
> Good to know. I did not see any wild regressions either and I do not 
> take Phoronix's reviews very seriously, but I was curious.

If you read Phoronix's articles (not just their headlines), even they
admit that this is not really a 'regression'. It's a perfectly sensible
bug fix. PCIe Active State Power Management was found to cause crashes
on some hardware, so since 2.6.38 it's explicitly disabled when the
system BIOS says it's not supported and only enabled where the BIOS
explicitly claims support for it (which is few machines). It's clearly a
defensible decision to prioritize 'not crashing people's machines' over
'ideal PM out of the box', and describing this as a 'regression' is
hardly accurate.

You can enable ASPM on machines where it's disabled by default with a
kernel parameter, if you're happy to turn it off again yourself if it
causes crashes, and if it actually makes a noticeable difference on your
machine.

See
http://www.fewt.com/2011/09/about-kernel-30-power-regression-myth.html
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net