[ntp:questions] Re: NMEA refclock - PPS not working

Ry malayter at gmail.com
Tue May 16 14:41:45 UTC 2006


> It's a common trick in modern PCs.  I doubt if it causes any troubles.

As ntpd uses the cycle counter on most platforms to get resolution
better than the timer interrupt, wouldn't frequency instability in that
counter cause ntpd accuracy issues?

> Note that the clock only gets stretched.  If you shrink it you violate
> timings.  It's pretty easy to measure.  Just compare the cycle counter
> with the clock and notice that your clock is running a bit slower than
> expected.

My reading of resources on the web seems to indicate that the frequency
spreading is equal in both directions, on the order of 0.5% (5000 ppm).
Support forum posts also seem to indicate that the spread specrum
option modulates the clocks on the AGP & PCI buses as well, depending
on the chipset & motherboard.

Are you saying this intentional frequency instability "averages out"
over time, letting NTPd's tick adjustments  (which are fairly static in
relation to the cycle counter's timescale) keep the clock stable?




More information about the questions mailing list