[time-nuts] pc clock crystal long term variations

Poul-Henning Kamp phk at phk.freebsd.dk
Sun Aug 8 21:40:35 UTC 2010

In message <20100808230143.2285e822.attila at kinali.ch>, Attila Kinali writes:

>What surprises me though is the stability of the crystal.
>In the months between reboots, there is less than +-1ppm variation,

>The machine is a 16y old DEC PC, which has been used as a desktop

DEC used way better crystals than the average PC-HW producers, so
I am not very surprised.  Not sure if this was just general good
engineering or a panicy reaction to all the trouble the alphas
gave them.

They had a few machines that could slot either x86 of alpha CPU
boards, and they were built from very high quality parts.

>And why doesnt the
>crystal relax back into it's old ppm value, but stays where it
>is after a reboot/power cylce?

You kernel tries to measure/estimate the crystal frequency at boot,
it does not get the same value on every boot.

Check your syslog/dmesg for the frequency estimate and you can see
this clearly.

Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

More information about the time-nuts mailing list