[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Time drift in amd64



On Thu, 03 Mar 2005 16:32:28 -0800
Alex Perry <alex.perry@qm.com> wrote:

> Carsten Prieß wrote:
> >> On Thu, 03 Mar 2005 18:45:45 -0500
> >> Peter Nelson <rufus@hackish.org> wrote:
> >>
> >>
> >>> Bharath Ramesh wrote:
> >>>
> >>>
> >>>> I just installed a fresh installation of amd64 on another athlon
> >box. >>> I find that the clock seems to be drifting by 8-10 seconds
> >every day. >>> Any idea how I can fix this drift. This drift seems to
> >be large. >>>
> >>>
> >>> Just install ntp-simple and it'll keep you within a few
> >milliseconds >> off  official atomic time.  If it asks you were to
> >point use >> pool.ntp.org.
> >>
> >>
> >> However that are only workarouds? What's with Clients with Dial-up
> >or > without I-Net connection?
> 
> Once you know what the drift value is, you can tell the kernel (using
> the adjtime related commands) and it will apply the correction for
> you.
I had the problem with x86. Not 8-10/s/h, but 10-15min/h. 
However the difference isn't constant. I solved the problem with ntp,
too ... but in my opinion this isn't clean.

> You do not need an always-on connection to use NTP.  It will simply
> take advantage of those times where the internet is available to get
> corrective data and use the most recently computed drift otherwise.
What's about greater and unconstant differences? min/h?

> If you don't have a valid connection at any time, and don't like
> setting the clock by hand every week or so, you can get a radio clock
> or a GPS and plug them in.  The "ntp" package (not ntp-simple) can use
> those.
Well, min/h and who has a GPS- or a radio reciever instead of or in
additon to an internet connection? Maybe EPG can help, but ...

Carsten



Reply to: