I'm astounded at the inaccuracy of the crystal clock in my PC, worst
i've ever seen at about 1 minute gain a day. My cheapish Swatch wrist
watch doesn't need reseting between summer and winter time changes and
thats about every 6 months. It just seems spot on.

Ignoramus30509 wrote:

A week ago I posted a surprised message about a $6 watch from walmart
that seemed to be accurate.

I have a little more data now.

About 2 weeks ago, I set this watch very accurately to time that is
kept by syncronizing with an NTP server (atomic clock), to the second.

Today I checked time again whiel trying to be very good at catching
the right moment. I typed "date" and pressed ENTER just as the arm of
the watch passed :00:00.

The result is that the watch is not even by one second off!!!
Obviously, there are limits to my own precision in how I pressed the
return key right when the watch ticked :00, plus the OS delay in
starting "date", but in any case I could not detect any difference.

I find it rather amazing, really.

The watch is a quartz watch with hands, "water resistant".

i

David Billington wrote:
I'm astounded at the inaccuracy of the crystal clock in my PC, worst
i've ever seen at about 1 minute gain a day. My cheapish Swatch wrist
watch doesn't need reseting between summer and winter time changes and
thats about every 6 months. It just seems spot on.

nah!, its quite normal for a pc clock to be way out, use a ntp program
that can sync the pc clock to an atomic clock, then you won't have problems.

I run NTP occasionally to syncronise but my main point was that my pc
keeps seriously crap time. Most pcs I have used keep to within seconds a
week if not months whereas my current pc of the last 6 years gains about
1 minute a day.

Ignoramus30509 wrote:

On Fri, 17 Mar 2006 22:12:09 +0000, David Billington wrote:

I'm astounded at the inaccuracy of the crystal clock in my PC, worst
i've ever seen at about 1 minute gain a day. My cheapish Swatch wrist
watch doesn't need reseting between summer and winter time changes and
thats about every 6 months. It just seems spot on.


I simply run ntp every hour. That's the way computers are supposed to
be run.

i

Ignoramus30509 wrote:

A week ago I posted a surprised message about a $6 watch from walmart
that seemed to be accurate.

I have a little more data now.

About 2 weeks ago, I set this watch very accurately to time that is
kept by syncronizing with an NTP server (atomic clock), to the second.

Today I checked time again whiel trying to be very good at catching
the right moment. I typed "date" and pressed ENTER just as the arm of
the watch passed :00:00.

The result is that the watch is not even by one second off!!!
Obviously, there are limits to my own precision in how I pressed the
return key right when the watch ticked :00, plus the OS delay in
starting "date", but in any case I could not detect any difference.

I find it rather amazing, really.

The watch is a quartz watch with hands, "water resistant".

i

David Billington wrote:
I'm astounded at the inaccuracy of the crystal clock in my PC, worst
i've ever seen at about 1 minute gain a day. My cheapish Swatch wrist
watch doesn't need reseting between summer and winter time changes and
thats about every 6 months. It just seems spot on.

Are the clocks in older PCs quartz oscillators, or are they something
inferior? I remember them keeping very poor time. When I got a Sun
workstation I remember being pleasantly surprised that the clock kept
time. But then the Sun uses this "Timekeeper RAM" chip, which is quite
expensive in itself.

Chris

Ignoramus30509 wrote:

All PCs I saw were lousy at keeping time with their clocks.

That's because you are looking at consumer PCs. Servers do
a *lot* better.

-----------------------------------------------------------

Christopher Tidy wrote:

Are the clocks in older PCs quartz oscillators, or are they something
inferior?

Quartz oscillators. Some older ones have a variable capacitor so
that you can adjust the speed with a screwdriver.

The problem is:

[1] Cheap crystals. If you make a huge number of PCs, saving a few
pennies by buying the crystals that didn't come out quite right
on frequncy is tempting.

[2] Cheap crystals. Buying crystals that change frequency with
temperature also saves money.

[3] Temperature swings. The inside of the PC is hot when on,
cold when off.

[4] Cheap drivers. Some kinds of driver chips help a crystal
to hold frequency better tha others. Guess which kind costs
less?

Linux users have a good solution to the problem of having an
inaccurate clock that is always fast or always slow. Whenever
ntpd starts it checks the frequency file (/etc/ntp/drift)
containing an estimate of clock frequency error and corrects
for it.