Jürgen Keil wrote:
I noticed a similar issue on my Tecra M5 when going from builds 109 to
builds 110 and 111. powertop would say that the cpu would be running at
1GHz but kstat would stay it was still at 1830000000
I stumbled upon an internal Solaris Nevada bug (CR# 6818514) that was
created for ATOM processors showing the same thing that suggests using:
cpupm enable poll-mode
Setting this on my M5 seems to cause the CPU to cycle down as expected.
Hmm, my ASUS N4L-VM DH Core2Duo box shows the same.
Seems to be a feature (or a bug?) of the new
"Power Aware Dispatcher" (PAD) support that
was added in build 110.
http://www.opensolaris.org/os/project/tesla/Work/CPUPM/
Maybe with the new PAD support (cpupm enabled in
"event-mode") the kernel is able to switch back to the
highest cpu frequency *much* faster (=> event driven),
so that "kstat cpu_info:::current_clock_Hz" now always
reports the highest cpu frequency?
(That is, monitoring the kstat cpu_info
current_clock_Hz isn't valid any more to find out
if cpu power management is working -- use powertop
instead)
I wasn't aware of PAD. Looking at powertop, I did see that the CPU
frequency was switching there, though it was always at the max according
to kstat (even when sitting completely idle). What you suggest could
very well be the case.
I just happened to notice it since my laptops fan was running more
often. I guess that I'll have to monitor how long the battery lasts
between the different modes.
-Phil
_______________________________________________
opensolaris-discuss mailing list
[email protected]
