DPS8/70M is a descendant of the 645. The /70 is a particular model in the line (speed, memory size, etc.) The M is the Mulitx suffix.

After the "merger" when Honeywell bought the GE computer dept, the Phoenix operation became Honeywell Large Information Systems Division, and they brought out a medium-scale integrated version of the GE 600 line as the Honeywell 6000 line. The Multics machine in this line was the Honeywell 6180, an improvement over the 645 in many areas. Subsequent Multics machines used the same processor architecture, with various master-mode-only tweaks to accommodate I/O evolution. But all were 36-bit, 8 index-register, A and Q register machines with many instructions in common.

One thing that changed in the 6180 and its descendants was that the GIOC was eliminated. The 6180 used a standard GCOS IOM plus a Datanet-30 descendant called a Datanet-355. Later Multics machines such as the DPS8/70M used other I/O arrangements inherited from GCOS. If you follow the Multics links page to the History of Bull, you'll find a French view of the subsequent history, and some information on John Couleur's disastrous "New System Architecture" which wasted many millions on features that were never used by GCOS and which caused the demise of Multics.

It could run Multics.

See System Design of a Computer for Time Sharing Applications and Multics Home

from Tom Van Vleck Jan 2015

Virtual Memory was not invented in New York. Here is the story as I understand it, with references. There are several techniques for providing virtual memory in a computer system: swapping and paging are the most common. (Segmentation is a hardware feature that can be used to assist paging or can be used as a program structuring principle.) Bob Daley always says that MIT's CTSS (1961-1973) was the first virtual machine software, since it provided a virtual IBM 7094 for use by batch programs. CTSS used swapping to provide virtual memory. http://multicians.org/thvv/7094.html MIT wanted to build a second generation time-sharing system with paged/segmented memory, and got millions (back when that was a lot of money) from ARPA to do so. This was Project MAC (1963-). http://multicians.org/project-mac.html Project MAC asked manufacturers for proposals. IBM's response was dismissive. They did not understand segmentation. MAC/BTL chose General Electric for the Multics system. http://multicians.org/history.html Some MIT employees left the Computation Center and went to work at IBM Scientific Center in Cambridge, including the late Bob Creasy. They invented virtual 360 computers provided by CP/67. http://multicians.org/thvv/360-67.html Perhaps the oral history you refer to is Melinda Varian's outstanding history of VM. (Melinda's husband Lee is a Multician.) http://www.leeandmelindavarian.com/Melinda/neuvm.pdf CP-67 and CMS were Type III software, unsupported. When IBM woke up to the need for time-sharing, it started a human wave design effort in Yorktown, with about 1000 programmers. TSS had severe performance and stability problems. TSS was announced in 1965, decomitted in 1968, un-decomitted in 1969, finally decommitted in 1971. [Varian] An academic paper on IBM time-sharing development was based on internal IBM documents made available as part of court cases against IBM. (The paper overstates the influence of TSS, since a fair number of 360/67 machines were listed as TSS machines in internal IBM documents, but in fact never ran anything but CP/CMS. IBM's internal in-fighting and competition was vicious, see Varian.) O'Neill, Judy E., 'Prestige Luster' and 'Snow-Balling Effects': IBM's Development of Computer Time-Sharing, IEEE Annals of the History of Computing Vol. 17, No. 2: Summer 1995, pp. 50-54. 1995. (I worked on CTSS and Multics at MIT and managed a 360/67 running CP/CMS in the 60s. I knew the developers of all three systems personally.) I remember the Thing King document. Web searches say it was written by Jeff Berryman at UBC about 1972. Jeff was a system programmer on UBC's 360/67, which ran the Michigan Terminal System, a system developed at Michigan when TSS was unsatisfactory and late. http://www.clock.org/~jss/work/mts/ There was strong resistance to virtual memory from traditional batch users, who thought it was inefficient. I remember a SHARE button that said "Will Virtual Memory be IBM's Watergate?" ... so that had to be 1972 or later.

In an attempt at humor - one could say
" This particular DPS8/70M was born on a little ranch in Arizona and sold to the no-talk folks out by Friendship Airport, worked hard all its life and was retired, given to the NCM, loaned to CHM. "

• Multics Home
• Multics Hardware Features
• System Design of a Computer for Time Sharing Applications by E. L. Glaser and J. F. Couleur and G. A. Oliver
• Multics CPU manual, with the whole instruction set from http://www.multicians.org/AL39.pdf - 1.5 MBytes
• "GE-635 Reference Manual" (preview version this site) from the transcriber Hans B Pufal
  {The GE 625/635 instructon set was a non-virtual memory subset of the GE 645 and Dockmaster}