[ntp:questions] May 2013 Windows Update causing rapid 'ntpd' drift

David Taylor david-taylor at blueyonder.co.uk.invalid
Sun May 19 07:07:45 UTC 2013


On 18/05/2013 14:50, starlight.2013q2 at binnacle.cx wrote:
> Hello,
>
> Just installed this month's Windows Update
> patches on an old/slow Windows XP laptop
> (1.5GHz Pentium-M) and it appears to have
> damaged Windows 'ntpd' time-keeping.
>
> Has anyone else observed this?
>
> Before the patch 'ntpd' frequency was -15.761
> PPM.  After the patches 'ntpd' frequency
> went to -500 PPM and the system was still
> gaining time so fast that 'ntpd' reset the
> time between -1.5 and -5 seconds every fifteen
> minutes.  This went on for 12 hours after the
> system was rebooted.  No change in the running
> applications and very light load most of the
> time.
>
> Restarted 'ntpd' with the same -g parameter
> and the problem persisted.
>
> Restarted 'ntpd' with the -M option added and it
> looks like it's doing ok now at around -55 PPM,
> but -M has its down-sides so I prefer to avoid
> using it if possible.
>
> Running a somewhat older version 4.2.4p4
> against two local CDMA time servers.  Generally
> get offsets under 100 microseconds.
> The version running here is patched to allow
> a one-second polling interval, which I've
> found necessary to obtain decent time-keeping
> with Windows.
>
> My best guess is that the problem results from
> kernel-mode 'Win32k.sys' driver patch
> KB2829361 aka MS13-046
>
>     http://support.microsoft.com/default.aspx?kbid=2829361
>
> Installed three other patches at the same time
> but two are for IE and one is for Active-X
> and don't seem likely culprits.  In case
> someone suspects otherwise, they are
> KB2847204, KB2829530, KB2820197.  The other
> May patches were installed three days ago
> since rebooting was not required for them.
>
> Any thoughts or comments are appreciated.
>
> Thanks

I have not seen the problem here, although my own 1.6 GHz Pentium-M PC 
is now running Windows-7, even though it has only 1.25 GB RAM.

On my XP systems, I do run with -M and have found no drawbacks.  The 
LAN-synched PCs run with a poll interval of either 32 or 64 seconds, and 
the performance can be seen here:

   http://www.satsignal.eu/mrtg/performance_ntp.php

Windows-8 should offer the chance of better time-keeping performance due 
to the introduction of a more precise GetSystemTime call which NTP uses 
(in the development branch).  4.2.4 is rather old, and it may be worth 
at least trying a more recent version.  I have some pre-compiled ones here:

   http://www.satsignal.eu/ntp/x86/

-- 
Cheers,
David
Web: http://www.satsignal.eu



More information about the questions mailing list