Messages in this thread | | | Date | Thu, 17 Apr 2008 11:03:04 -0700 (PDT) | From | Marc Perkel <> | Subject | Re: AMD Quad Core clock problem? |
| |
--- Bart Van Assche <bart.vanassche@gmail.com> wrote:
> On Wed, Apr 16, 2008 at 11:44 PM, Marc Perkel > <mperkel@yahoo.com> wrote: > > --- Bart Van Assche <bart.vanassche@gmail.com> > wrote: > > > It's better to correct the clock frequency once > with > > > the tickadj or adjtimex, so you don't have to > add > > > ntpdate to cron and you can still run ntpd. > > > > > > Bart. > > > > ok - thanks. How do I do that? > > You can do that as follows: > * First of all, make sure that adjtimex is > installed. Some distro's > include this command in util-linux. > * Query the current tick value via adjtimex -p | > grep tick. E.g. with > HZ = 100, tick == 10000. > * Multiply the tick value with the relative clock > error: when e.g. 3s > are lost every hour, the correct tick value is 10000 > * (1 + 3/3600) = > 10008. > * Pass this new tick value to the kernel via > adjtimex -t 10008. > * Stop and restart ntpd such that it forgets any > previous frequency estimates. > * Keep an eye on the output of adjtimex -p|grep > frequency to see > whether the frequency estimate converges (should be > stable within 10 > ppm after an hour. See also man adjtimex for the > units of this value). > > Bart. >
Thanks Bart - that seems to have worked. A value of 10025 seems to get it close enough for NTPD to lock on. I'm using tickadj.
So - the question is - is this a kernel bug not calibrating quad core phenoms right or do I just have a bad motherboard?
Marc Perkel Junk Email Filter dot com http://www.junkemailfilter.com
____________________________________________________________________________________ Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
| |