[ntp:questions] Sub-millisecond NTP synchronization for local network

Hal Murray hal-usenet at ip-64-139-1-69.sjc.megapath.net
Sun Dec 7 23:29:19 UTC 2008


>Obviously using something like the number of cpu cycles is the wrong thing
>to use for a clock if the computer is to act as a good clock. HPET should
>solve that if only they got it set up properly.

What's wrong with using the CPU clock as a clock?  Why is HPET any better?

I know of a couple of quirks with the CPU clock.

One is that is typically fuzzed by a fraction of a percent to
spread the EMI around to get past FCC regulations.  That results
in a number that is reasonably constant. It's just not the number
you expect.  For example, this PC is running at 564.775 MHz rather
than 566 (which is probably 566.666).

The other is that modern CPUs switch the clock speed to control
power and temperature.  I'm not familiar with the details.  I expect
that wouldn't cause much of a problem if it could be coordinated
with the time keeping.

-- 
These are my opinions, not necessarily my employer's.  I hate spam.




More information about the questions mailing list