On St, 2008-08-13 at 15:05 +0200, Pavel Machek wrote:
Hello,
I apologize, I have not seen your reply, the newest update already
describes my situation.
I tried nohz=off; result was stably working system, but the cause
could be that it virtually did not enter any of deeper C states (I also
had HZ configured to 1000, but I'm not sure whether that matters when
calculating efficieny of switching to Cx state). I also tried to disable
hpet at all, and to disable support for high resolution timers in
kernel, but result was the same, IIRC. Unless I wasn't entering C2 or
C3, the system was stable.
Milan
--