I wonder if the clock has a broken quartz crystal ? Has the
computer ever been dropped or kicked ?
When the computer is running and the OS is booted, the time is
maintained by incrementing a variable stored in system RAM. In
other words, the OS uses a "software clock" and stops looking
at the RTC. The RTC is too slow, to be read regularly while
the computer is operating. The hardware interface is prehistoric.
When the system is about to shut down, the time is written back
to the RTC (or it can be).
The RTC contains (at least portions of it), are a ripple counter.
The value strobed into the flip flops would stay put, if the
32768 Hz clock generator signal was lost. Then, the next time
you boot the computer, and the OS reads the RTC and copies the
value into system memory, it's going to be reading the pattern
in those flip flops.
So your symptoms are consistent with a failed 32768 Hz quartz crystal
or the oscillator it is connected to. If the same value written
at the end of the day, is showing up the next morning, there there
are no pulses coming from the 32768 Hz source.
(too much emphasis on the older implementations here...)
http://en.wikipedia.org/wiki/Real-time_clock
The RTC and CMOS 256 byte RAM block, are powered by the CMOS battery.
If there was a total loss of power there, then the RTC would not work,
and the BIOS would forget key settings. On pre-built computers (like
my laptop from last year), I can only see one setting in the
BIOS, so there really isn't much to store in there any more. I'd
be hard pressed, to detect a change in the BIOS settings caused
by a battery failure.
If the RTC loses power, the BIOS initializes the flip flips to a well
known date and time. On older RTC designs, this might have been the
year 1970. But more modern computers, have been using "magic date"
values a bit above the year 2000.
In any case, if you saw a nice round value like Jan.1 1970, then that
would be telling you the RTC flip flops were uninitialized and had
been programmed by the BIOS. If the time matched the time at which
the computer was shut down, that tells me the RTC is powered all night,
but there is nothing feeding clock pulses to it. With no 32768 Hz
pulses, the flip flop based counter cannot advance.
The device on the left here, is a 32768, based on size and form factor.
They're used in digital watches. And you can find one next to the
Southbridge (where it is used by the Real Time Clock or RTC). On the
computer, the RTC draws about 10 microamps of current, while on a
digital watch, the figure can be as low as 2 microamps.
http://www.softwareforeducation.com/electronics/Circuits/QuartzCrysta...
The circuit it connects to, looks like what is connected to pin 12 and 13
in this diagram. A quartz crystal has two legs, and there are a couple
small capacitors (measured in picofarads) in the circuit as well.
http://www.moty22.co.uk/img/lcd_counter.gif
Since the active part is inside a tin can, you can't look at it.
An ohmmeter would not be of much use either.
In practical terms, I'd expect to have to replace the motherboard.
On the one hand, you could probably dig up another crystal, but
there are a whole page of specs that have to match. At work, I've
always sought the assistance of a "quartz crystal expert", when
fiddling with those. Nothing bad would happen if you used the
wrong one, but perhaps it would fail again after some period
of time (if overdriven).
If you want, you can visually inspect the motherboard, and look for
the tin can and the two caps. They should be near the Southbridge,
and it isn't a good idea for the motherboard designer to put them
half way across the board. They should stay close to the equivalent
of pin 12 and pin 13 in that sample diagram, for best results.