When did computer clocks get so bad?
Hi there,
I have a question and I cannot think of anyone better to ask, for a
project we need to time some things which are connected to a computer,
using NTP and later using a GPS over bluetooth serial ports, we have
discovered that computer clocks are terrible
If you remove a linear drift (for example assuming it ticks at 1.00026
seconds per second) it gets less terrible, and Linux can do this but it
is clear that the computer clock doesn't expose this coefficient to the
OS to let it compensate, it must be found (eg through NTP) - any ideas why?
But more concretely, my watch is actually pretty good, it's off by < 3
seconds and hasn't been set probably this year (I don't tend to bother
with DST stuff, not for any reason just never get round to it) - when I
was growing up and even now wall-clocks are not so terrible that I have
to fix them (or NTP does with computers) very routinely.
My theory is that super cheap crappy quartz clocks are now used in
things which can be reasonably expected to be online most of the time,
and thus use NTP - my watch cannot (and probably has temperature
correction too? Given the varied temps it is exposed to) any truth to this?
This is a very open ended question I understand, but if clocks were as
terrible as I've found every computer and thing I've checked recently,
why don't I remember setting wall clocks easily once a week?
Alec
Alec Teal writes:
My theory is that super cheap crappy quartz clocks are now used in
things which can be reasonably expected to be online most of the time,
There are multiple answers to your question.
The funny one is: When they set fire to a prototype motherboard
at Intel Architecture Labs.
Unfortunately that is not my story to tell.
But for the PC-ecosystem, that really is the answer:
When Intel had to start modulating clock-frequency in order to not
set things on fire.
Their execution was far from stellar, because they had their eyes
only on Windows, which for all intents and purposes had no timekeeping
worth anything at the time.
By now they have it relatively under control, and due to the very
steep PLL multipliers high end kit actually have comparatively
good XTALS in order to keep the jitter within spec.
For the Internet of Shit segment, the answer is it was never any
good to begin with.
Since these embedded chips generally are incredibly robust with
respect to timing, the xtal on the BOM is the cheapest that will
meet spec.
The one saving grace is external high-speed interfaces like USB-3
and 10G ethernet: You need good-ish xtals before it even works.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@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.
Alec,
I seem to perceive that PC clocks have gotten quite a bit better since
their early days. What do others think?
Dana
On Wed, Sep 29, 2021 at 3:10 PM Alec Teal alec@unifiedmathematics.com
wrote:
Hi there,
I have a question and I cannot think of anyone better to ask, for a
project we need to time some things which are connected to a computer,
using NTP and later using a GPS over bluetooth serial ports, we have
discovered that computer clocks are terrible
If you remove a linear drift (for example assuming it ticks at 1.00026
seconds per second) it gets less terrible, and Linux can do this but it
is clear that the computer clock doesn't expose this coefficient to the
OS to let it compensate, it must be found (eg through NTP) - any ideas why?
But more concretely, my watch is actually pretty good, it's off by < 3
seconds and hasn't been set probably this year (I don't tend to bother
with DST stuff, not for any reason just never get round to it) - when I
was growing up and even now wall-clocks are not so terrible that I have
to fix them (or NTP does with computers) very routinely.
My theory is that super cheap crappy quartz clocks are now used in
things which can be reasonably expected to be online most of the time,
and thus use NTP - my watch cannot (and probably has temperature
correction too? Given the varied temps it is exposed to) any truth to this?
This is a very open ended question I understand, but if clocks were as
terrible as I've found every computer and thing I've checked recently,
why don't I remember setting wall clocks easily once a week?
Alec
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send
an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
I assume you're talking about the RTC chip that runs off the CMOS BIOS
memory settings battery, to keep time whether the computer is on or off.
These are as you suspect, typically cheap items that get the basic job
done. You may be able to find better grade ones. It also depends on
whether the timing crystal is built into the IC, or a separate piece. In
this case, you could fairly easily hack in a reference that's as good as
you want. Usually it's a 32.768 kHz "watch" type resonator, and
disconnecting one end and hooking in an external source will drive it
instead. If you provide the same nominal frequency, that's much more
accurate and stable, can run the RTC under all conditions, and is always
on, then it should keep time as good as your reference.
Ed
On Wed, Sep 29, 2021 at 1:48 PM Poul-Henning Kamp phk@phk.freebsd.dk wrote:
Since these embedded chips generally are incredibly robust with
respect to timing, the xtal on the BOM is the cheapest that will
meet spec.
Crystals? We don't need no stinking crystals!
At least for the RTC. Chips like, e.g. the NXP iMX6, can use an
internal ring oscillator in place of the customary 32kHz crystal to
save even more on BOM cost and board space. Still needs a faster xtal
for full operation, but low power mode and RTC can use the ring
oscillator and keep the 24 MHz xtal powered down.
Hi
If your wrist watch is holding 3 seconds per year without some sort of external
correction, that’s pretty amazing …. 10 seconds a month is doing well.
Bob
On Sep 29, 2021, at 2:35 PM, Alec Teal alec@unifiedmathematics.com wrote:
Hi there,
I have a question and I cannot think of anyone better to ask, for a project we need to time some things which are connected to a computer, using NTP and later using a GPS over bluetooth serial ports, we have discovered that computer clocks are terrible
If you remove a linear drift (for example assuming it ticks at 1.00026 seconds per second) it gets less terrible, and Linux can do this but it is clear that the computer clock doesn't expose this coefficient to the OS to let it compensate, it must be found (eg through NTP) - any ideas why?
But more concretely, my watch is actually pretty good, it's off by < 3 seconds and hasn't been set probably this year (I don't tend to bother with DST stuff, not for any reason just never get round to it) - when I was growing up and even now wall-clocks are not so terrible that I have to fix them (or NTP does with computers) very routinely.
My theory is that super cheap crappy quartz clocks are now used in things which can be reasonably expected to be online most of the time, and thus use NTP - my watch cannot (and probably has temperature correction too? Given the varied temps it is exposed to) any truth to this?
This is a very open ended question I understand, but if clocks were as terrible as I've found every computer and thing I've checked recently, why don't I remember setting wall clocks easily once a week?
Alec
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
My Dad who taught Electrical Engineering told me that when parts are sold
in different quality grades (like resistors or in this case, crystals),
that the parts that meet the highest spec are sold for a high price, then
the next highest quality spec parts are sold for less, and so on. So your
typical 30ppm computer chip will probably have a bias that is either about
+25ppm or about -25ppm, but definitely NOT likely to be in the range -10ppm
to +10ppm.
Cheers,
James
On Wed, Sep 29, 2021 at 1:10 PM Alec Teal alec@unifiedmathematics.com
wrote:
Hi there,
I have a question and I cannot think of anyone better to ask, for a
project we need to time some things which are connected to a computer,
using NTP and later using a GPS over bluetooth serial ports, we have
discovered that computer clocks are terrible
If you remove a linear drift (for example assuming it ticks at 1.00026
seconds per second) it gets less terrible, and Linux can do this but it
is clear that the computer clock doesn't expose this coefficient to the
OS to let it compensate, it must be found (eg through NTP) - any ideas why?
But more concretely, my watch is actually pretty good, it's off by < 3
seconds and hasn't been set probably this year (I don't tend to bother
with DST stuff, not for any reason just never get round to it) - when I
was growing up and even now wall-clocks are not so terrible that I have
to fix them (or NTP does with computers) very routinely.
My theory is that super cheap crappy quartz clocks are now used in
things which can be reasonably expected to be online most of the time,
and thus use NTP - my watch cannot (and probably has temperature
correction too? Given the varied temps it is exposed to) any truth to this?
This is a very open ended question I understand, but if clocks were as
terrible as I've found every computer and thing I've checked recently,
why don't I remember setting wall clocks easily once a week?
Alec
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send
an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
--
James Perkins james@loowit.net KN1X www.loowit.net/~james
2030 W 28th Ave, Eugene OR 97405 +1.971.344.3969 mobile
Alternate email: opalmirror@gmail.com