The 360/85 had the microcode for its base instruction set in ROS (ROM)
so it would not have needed a device like the 23FD to get started. It
was also shipped starting in December 1969 which is before the date the
23FD is reputed to have been ready. A paper I found at
https://dl.acm.org/doi/epdf/10.1145/1476936.1476969 describes
instructions in the the 360/85 to load words into the writeable control
store from main memory. The 370 systems on the other hand had no
control store in ROS so needed a device to do the Initial Microcode Load
(IML) to get the machine of the ground and I suspect the 23FD and later
IML microcode sources where just part of the IML system as something
would have to stuff the microcode into the control store. I see some
370 documents mention a "Service Processor" that loads the microcode,
one candidate for this service processor would be one of the family of
Universal Controllers that IBM used in a wide variety of machines dating
back to the early to mid 1970s.
Paul.
On 2026-06-09 20:21, Jon Elson via cctalk wrote:
On 6/9/26 16:24, Tom Gardner via cctalk wrote:
Lot's of mis-information which I think I correct below
I'm pretty sure the first shipments of the 23FD was in January 1971
the with 2835 Storage Control Unit for the 2305 Fixed Head DASD and
the S/370 M155; it might have been simultaneous but the 2835 did
attach to existing S/360 M85 2880 Block Multiplexer channel and the
M85 shipped in December 1969. It was followed later that year with
the shipment in the S/370 M165 and then the rest of the S/370. FWIW,
the 2835 SCU was very similar to the 3380 SCU, mainly differing due
to the parallel data transfer from the drive. The 23FD was a read
only device
The 360/85 was EXTREMELY close to the 370/165, in fact it was the
prototype of the 165. Looking through principles of operation from
the two models shows instruction timings are totally identical. The
model 165 had a 23FD to load microcode and diagnostics. The 360/85
had read-only 360 microcode, but emulators and diagnostics were
loadable. I can only assume that was done via a 23FD. The major
difference between 360/85 and 370/165 hardware was that microcode was
partially read-only in the 85, with 500 words or writeable control
store, implemented in 16-bit MST4 chips. The 165 had all microcode in
writable ECL static RAM in 64-bit chips. The /85 cache (storage
buffer) was implemented in the same 16-bit chip technology, the 165
was done with the 64-bit chips. There were some slight differences in
the 360 vs. 370 processor status word, also.
Jon
