[ntp:questions] NTPD can take 10 hours to achieve stability

unruh unruh at wormhole.physics.ubc.ca
Tue Apr 19 06:03:58 UTC 2011

On 2011-04-18, Chris Albertson <albertson.chris at gmail.com> wrote:
> On Mon, Apr 18, 2011 at 3:49 PM, unruh <unruh at wormhole.physics.ubc.ca> wrote:
>>> Thanks for all the great answers. ?Now for a harder question, how does the accuracy of the local clock source
>>> affect the accuracy of ntpd.
>> What local clock source?
> That is a good question:  "Which clock?"  One clock might be the
> reference clock for a stratum 1 NTP server.  This is typically a GPS
> receiver and these vary a LOT.  Some can supply a pulse per second
> with 1 sigma error of only a hand full of nanoseconds and other GPSes
> have hundred of microseconds of error.  That is a factor of over
> 10,000 difference in accuracy.

That sounds like the difference between a PPS gps receiver or one whith
simple NMEA output. The latter should NOT be considered a stratum 0

> The other clock is the crystal inside the computer that NTP tries to
> "discipline" by using the above reference clock.  I think what matters
> most is "stability" rather than "accuracy".   In other words it is
> fine if this oscillator runs at the wrong speed but not good if the
> speed is not constant with respect to time.  The best oscillators are
> housed in temperature controlled ovens but these are not common inside
> PCs.

What is worse is that modern cpus in order to conserver energy keep
changing their rate, which makes them really difficult to use as a good


More information about the questions mailing list