Delete comment from: Ken Shirriff's blog
Looking back through earlier blog entries I came across one "TROS: How IBM mainframes stored microcode in transformers". There are a few paragraphs there which show the 360/65 and 67 used BCROS (balanced capacitor read only storage) and explains where the SaranWrap came into the picture. I seem to remember the FE was Kent Pritchard working out of the Bridgeport CT IBM office. A wonderfully dedicated team of FE's installed the 360/67 in a partially completed (unheated) computer centre in the winter of '69/70. They worked heated by gas/kerosene space heaters... periodically heading out to the parking lot when the fumes became too much. I understood that the machine in question (a duplex with 2 CPUs) had come from the Ames Airforce base in CA. We initially configured it (a big switching unit) as two separate "simplex" machines and then sat down to write upgrades to our OS to allow "duplex" operation. One copy of the OS controlling 2 CPU's and freeing up memory previously used by the 2nd copy of the OS. Memory treated as one shared resource. Memory was expensive, scarce and temperamental. Initially a total of 8MB in temperature controlled 2361 units (2MB per 4'x 6'x 8' box). Brilliant AMS engineers came up with plug compatible semiconductor memory using in house newly developed chips (I've no idea of the chip number). Suddenly we could afford to put on more memory units (to 16MB) and run close on 200 concurrent customers thanks to 2301 drums and VIRTUAL MEMORY (initially developed as "one level storage" by a team at Ferranti and Manchester University around 1960). The era of Time Sharing died around a decade later with the arrival of the IBM PC in 1981... Vita brevis.
Feb 2, 2021, 7:44:02 PM
Posted to IBM paperweight teardown: Reverse-engineering 1970s memory chips