[ntp:questions] NTP Third party client on Windows OS ?

Heiko Gerstung heiko_removeme_.gerstung at meinberg.de
Wed Sep 19 12:31:33 UTC 2007


Evandro Menezes schrieb:
> Martin,
> 
> First off, I appreciate your interest.
> 
> On Sep 18, 3:53 am, Martin Burnicki <martin.burni... at meinberg.de>
> wrote:
>> I would be interested in how you found that out, what version of w32time and
>> NTP you were comparing, and what the differences are.
> 
> I run Windows XP with W32TIME 5.1.2600.2180 and I tried NTP 4.2.4p3.
> 
>> Last time I compared the Windows system which was diciplined by either
>> w32time, or NTP, to a reference time provided by one of our GPS cards the
>> NTP program provided a much better accuracy than w32time.
> 
> Mind you, I don't have an accurate clock source, so I sync up to
> pool.ntp.org servers and to company-wide servers (also running NTP).
> 
> What I see is that with NTP the time drifts by .128s quite quickly and
> as soon as the stepout limit of 900s expires, time is stepped by
> about .5s.

I see the same behavior on my old laptop when I enable the multimedia-timer 
switch of ntpd (-M). If I turn the MMTimer feature off, it is stable.

> It seems to me that my system clock sucks, but yet if use W32TIME,
> even if it starts over 1s off, it gets in sync in about 30min, keeping
> the difference between it and its reference within 50ms.
> 
> Perhaps the PLL algorithms of NTP can't cope with such a lousy clock,
> perhaps the Windows kernel doesn't allow for a fine timing control, I
> don't really know.  However, as much a fan of NTP as I am, it seems
> that my system is happier with W32TIME.
> 
> I've been running NTP on it for almost 24h now along with your Time
> Server Monitor program and the charts look horrible.  Something like a
> saw-tooth...

With the Time Server Monitor application you can easily switch off the MMTimer 
feature of NTPD (see the "Settings" button on the "NTP Service" tab, uncheck the 
"Set MMTimer to highest resolution" box).

Best Regards,
  Heiko


> 
> Thanks.
> 




More information about the questions mailing list