[ntp:questions] What level of timesynch error is typical onWinXP?

David J Taylor david-taylor at blueyonder.co.uk.invalid
Fri Oct 22 14:04:07 UTC 2010


"Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message 
news:HvGdnY984a8wClzRnZ2dnUVZ_tCdnZ2d at giganews.com...
[]
> AFAIK, *all* versions of NTPD require several hours to reach a stable 
> state and a close approximation to the correct time.  NTPD was designed 
> for operation 24x365.  Every time you shut down you will encounter a 
> multi-hour settling period.  The time will be *close* to correct within 
> thirty minutes or so; getting within microseconds takes a good deal 
> longer!
>
> The solution is obvious: don't shut down! If you are in a situation 
> where only 9-5, M-F operation is possible and microsecond accuracy is 
> required, NTPD is probably the wrong tool for the job.  See CHRONY.  Its 
> design goals are different and include fast startup.

Thanks, Richard.  I normally run 24x7 here on the critical PCs, so I was 
simply responding to Dave Hart's suggestion to test the later version.

  "4.2.7p52, p53, and p54 each had initial convergence improvements from 
Dr. Mills.  I encourage you to play with 4.2.7p54 or later and see for 
yourself.  ntpd now eliminates the majority of the startup offset in 5 
minutes (with a drift file) or 10 minutes (without).  During the startup 
period, the frequency adjustment used to slew away the startup offset is 
not retained, so there should be little "ringing"."

I had interpreted "initial convergence" as meaning that several hour 
period.

Chrony does not support Windows, and there are no plans to do so 
(according to 11.2 in their FAQ).  It might be a nice client for portable 
PCs, although for my purposes the existing NTP for Windows is quite 
adequate for use on portables.

Cheers,
David 




More information about the questions mailing list