There are a number of very significant projects that never came to market as originally envisioned, although the lessons learned sometimes got applied to later offerings.
Was to have used the KB-11Cm CPU modified to interlock the ASRB instruction in memory, allow cache bypassing on a per page basis, and tweaked powerfail/re-start for the 11/74. These were shipped with MKA-11 memory boxes as the 11/70mu system, all to AT&T, I believe. The only multiprocessing systems were Castor/Pollux in the RSX development group in ZK (Software Engineering at Spitbrook Road, Nashua NH), and the RSX DECnet group in Tewksbury MA. Ontario Hydro also had one for field test and invented "just say no" when the field test ended. :-) BTW, HRC, the reconfiguration task, is short for "Hercules", who tamed Cerberus (the MP version of TRAX/M-PLUS).
Probably the most famous PDP-11 that never was. Designed as a 4-way multiprocessor with four 11/70 processors clustered around shared memory. Configuration was SMP. Furthermore, the processors, which had been originally designed as FORTRAN crunchers, were to be equipped with the Commercial Instruction Set (CIS) for Cobol applications. Prototypes were built and a number of alpha units installed, but the product never came to market. Reasons for it never being shipped are various and may be apocryphal: the backplane would have been so complex to be almost unmanufacturable; the cost due to the complex backplane would have been too high; the 11/74 with CIS would have blown the new VAX-11/780 out of the water in commercial applications. Those that truly know are not talking :-)
The backplanes are almost standard KB-11C. The microcode is changed to interlock the ASRB instruction, cache bypassing is added and a few other minor changes. No problem to manufacture.
The memory is the MKA-11, with 4 memory ports. That's 32 flat gray cables in and out of each box except the last, plus 4 box controller cables and battery backup.
The CIS equipped unit was the KB-11E cpu, aka 11/110. Another contempory of these was the 11/68. (see Bluefin)
There's also the IIST, the DIP-11, and so on, for which no useful information has been found to include in this FAQ.
Robert Boers reports:
I was in DEC Engineering for more than 20 years. One piece of data I can add: The reason that the 11/74 was never produced was not related to competition with the 11/780. The reason was that Field Service wanted an exorbitant upfront amount from Engineering, if they had to lower the yearly maintenance fee for the 11/74 to a 'normal' % of the purchase price (around 10% if I remember correctly). DEC produced a few 100 of the 11/74, which were mainly used internally. I remember that they were quite reliable. I believe we had some near my office with a dual CPU, but I don't recall ever seen one in use with 4.
This was to have been the `real' /74. A KB-11Cm, cleaned up, and with CIS included. Canned just as it was to go into production. The maintenance prints for the CIS were in the hardware catalogue, as were the MKA-11 manuals. The KB-11Cm tech manual is listed as EK-KB11CM-TM, yep 2-6-2 not 2-5-2, in at least 2 places (see the Part Number Format section).
The /60 was to have been a 22 bitter originally. The 11/68 was to have had up to 16 cpus, each as powerful as a 4-cpu /74. Canned to prevent competition with a future 32 bit product.
18 bit system done at CMU.
The MKA-11 manuals refer to these. Perhaps they were going to be what became the 11/780 VAX?