[ntp:questions] NTP 4.2.6p4 Released

Dave Hart hart at ntp.org
Wed Sep 28 07:38:30 UTC 2011

On Wed, Sep 28, 2011 at 05:20 UTC, David J Taylor wrote:
>> * Added the NTPD_TICKADJ_PPM environment variable for specifying baseline
>>  clock slew on Microsoft Windows
> Does NTPD_TICKADJ_PPM allow the use of clocks which are more than 500 ppm
> out?  Sorry if I missed any earlier discussion on this (and I don't run
> Linux).  How does one go about determining the value to set - just
> guesswork?

Yes, it should, just as tickadj and similar capabilities do for POSIX
systems.  You can determine the value to set any way you like, from
measurement to guesswork to dart throws.  The adjustment is done at a
low level in the Windows ntpd port, invisible to the portable ntpd
code, but is expressed in PPM with the same sign ntpd uses.  If ntpd
is bouncing up against the +500 PPM limit occasionally, using a
positive NTPD_TICKADJ_PPM should improve the situation.

> I'm interested because someone has just reported problems with NTP on a
> system which gain (or loses) 1 second in 10 minutes!  1 part on 600!  This
> adjustment could be most helpful.

It might help, I hope so.  Hopefully the clock rate error is
consistent enough that after the extreme compensation, it stays well
under 500 PPM.

Dave Hart

More information about the questions mailing list