On 2010-02-18, wrote:
On Tue, 16 Feb 2010 17:37:23 -0800, Bruce L. Bergman
wrote:

[ ... ]

All well and good, but you'll be changing the CMOS battery more
often. Better than the old days, when the battery was molded into the
RTC chip, and you have to replace the whole thing.


That's not the "good old days" - that's the "in-between nasties". In
the good old days there was a big square CMos battery which could be
replaced with a 4-pack of AA dry-cells.
The self-powered RTC chip was, thankfully, a rather short term
aberation - or was that an abortion.

Actually -- that chip was something which Sun kept for *way* too
long. The clock chip actually is a TOD (Time Of Day) clock/calendar,
not a Real-Time Clock which generates precisely spaced pulses for task
switching in time sensitive applications like the CPU controlling a CNC
machine tool. The two terms have gotten mixed up over the years.

The one Sun used (from Toshiba, IIRC) had 2K of CMOS RAM in
addition to the clock and the potted-in cell on machines at least from
the early SPARC-1 machines up through the Ultra-60 in my experience.
The problem was that the RAM contained (in areas inaccessible for write
by the firmwre) the HOSTID and the MAC address (raw ethernet address).
And if you were using licensed software (some compilers and fancy backup
systems were examples which I know of -- probably some database programs
and CAD packages as well) depended on the HOSTID and the MAC address.
It was possible to re-write those sections using a fcode program (FORTH)
keyed into the NVRAM part, but it was a real pain. Each had a bar code
label and you could order a replacement from Sun by barcode number. The
rest of the 2K was used for the eeprom setup variables, and there was a
keyboard sequence during boot to reset the variables to the factory
defaults.

About the time of the SS-2, a company named Solbourne brought
out a semi-clone which had the HOSTID and MAC address in a bipolar PROM,
and the EEPROM and clock were backed by a coin cell which could be
easily replaced. The bipolar PROM was mounted in a socket (as was the
NVRAM/CLOCK chip on the SPARC systems), and if you had to replace a
system board, it came without the bipolar prom -- you were supposed to
swap it over to the new board so your licensed software would continue
to operate.

Yes -- I had to replace a system board in a SS-5 which was
running a licensed compiler, and so I swapped over the NVRAM/CLOCK chip
when I changed. (It helps to have a collection of spare older systems
sitting around when one dies. :-)

The Sun use of the clock/RAM chip started perhaps around 1980 or
so, and continued until the Sun Blade 1000/Sun Blade 2000/Sun Fire 280R
and later UltraSPARC systems, which put most of the information in a
SEEEPROM (Serial Electrically Erasable EPROM) and the clock is powered
by a coin cell in a holder on the system board (plus, of course, the
power supply itself when the system is plugged in. :-)

So -- that Toshiba chip was in service far too long in Sun
systems.


Enjoy,
DoN.

--
Email: | Voice (all times): (703) 938-4564
(too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html
--- Black Holes are where God is dividing by zero ---