Chris,
If this is the machine that I am thinking of, it was made by
Intel and marketed by them as the AS/5. Control Data sold it as the Omega
system, but I am not clear on the "480" nomenclature. It was equivalent to the
IBM 370/158, and not only was capable of running 370/158 operating code, but
diagnostics as well. The 370/158 was only available in max 3mb memory, and
Intel was marketing add-on memory, since the memory addressing bit was
available up to 8mb. So one of the selling points of the Omega was that it came
with 8mb of memory. One of the Omega frames was actually the same add-on memory
frame that Intel sold to IBM customers.
Now here is the kicker. The 370 mainframes had swing gates with motherboards,
and each motherboard had a couple of logic cards with components soldered on
them. The Intel box had some 40 or so large boards, each with several hundred
pluggable chips. In other words, they bypassed the logic card layer and went
from board to chip. So once every six months or so, the mainframe had to be
powered down, the boards had to be sprayed with a solvent, sit overnight, and
the next morning sprayed down with an alcohol aerosol to disburse the solvent.
This was required because the legs of the pluggable chips would begin to
oxidize over time, causing loose connections and timing problems and logic
failures.
The other kicker, is that you could run the 370/158 diagnostics, but the
diagnostics would tell you what the failure was ... if it was a 370/158. There
was no chart or graph that told you what card in the 370 corresponded exactly
with what board in the Omega. For instance, on the FIPS channels, all the bus
and tag drivers were on one board, and all the bus and tag receivers were on
another board. In the 158 all the channel 2 logic was on one board, you could
replace that one board and be done with it. The Omega was a bit different.
How to troubleshoot this beast, since you had to troubleshoot down to the chip
? The only spare boards normally stocked were the floating point arithmetic
boards, since there was no way to scope or test those. We took over maintenance
of five of these machines at the Ministry of PTT in Riyadh, Saudi Arabia in
1988. Our sheikh won the contract for maintaining the facility, and he had
people who could handle the monitors and peripherals, but not the mainframes.
So you could cycle the failing diagnostic microcode word and scope the signal
down to the failing chip. Or you could use the second level flowcharts, narrow
down the problem to a dozen or so chips, and then swap them to other parts of
the machine to break the failure down to one chip. Of course, because of the
design, you had issues with loose chips, loose legs, dirty legs, and even
broken legs that would cause intermittent failures. Loads of fun !!
In the 70s the 370/158 was the workhorse of the industry, so it was fun to run
into these old machines and work on them for a while. When you work on
something like this, you really get to see into the depths of the mainframe and
see exactly what it is doing.
Hope you have a great day !
Anthony B
IBM Bluemix
