IBM Advanced Computing Systems -- Draft Timeline

Mark Smotherman
last updated July 8, 2010; correction made on May 12, 2011

(This reflects my current understanding. I am sure there are errors. I very much appreciate corrections and new material. I also hope the style choices are helpful rather than distracting; there was beginning to be so much important contextual information that the actual ACS-related dates were being swamped.)

Project Timeline

(includes information about IBM in green italics to help set the context of ACS within the company)
(also includes information on competitors machines in red emphasis)

• 1953
  • April: first IBM 701 delivered

• 1954
  • Gene Amdahl, who joined IBM in 1952 after completing his Ph.D. in physics at Univ. of Wisconsin, leads the 701 redesign that results in the 704
  • August: Stephen Dunwell's first "Datatron" memo

• 1955
  • January: Livermore requests a supercomputer
  • April: IBM "8 to 10 Megacycle machine" proposed to Livermore; UNIVAC would win the sale with their LARC proposal
  • September: presentation to Los Alamos
  • October: IBM 704 and 705 announced
  • Gene Amdahl resigns from IBM "after losing a struggle for design control of the Stretch computer" to Dunwell [Pugh, et al., p. 123]

• 1956
  • January: official start of Stretch project
  • November: contract signed with Los Alamos for Stretch - $4.3M with delivery in 42 months

• 1957
  • January: Stretch transferred from Research to Development
  • Stretch designed as "three-in-one": Basic, Sigma, and Harvest
  • November: Cocke-Kolsky simulator first used

• 1958
  • June: IBM 7090 project started, using Stretch circuits, memories, and frames

• 1959
  • December: first deliveries of IBM 7090
  • December: presentation on Stretch at EJCC

• 1960
  • Amdahl returns to IBM, works in Research; led study team with John Cocke to find "literally a hundred" ways to improve the 7090 (in 1961?) [Pugh, et al., p. 123; Bashe]
  • several engineering improvements for Stretch considered under the named "SuperStretch"
  • November 2: meeting at Yorktown with Livermore representatives on computers beyond Stretch; E.A. Adams, W.L. Shevel, and A.L. Leiner made presentations; Livermore felt that a machine 10-20 times the Stretch "would be of considerable interest"

• 1961
  • January: Work on what will become Project X started within Research [Fisher, p. 154]; Max Paley convenes meeting on "Stretch - Where do we go from here?"
  • January: 8000 series proposal; model 8106 had been under construction in 1960 by DSD; team members included Fred Brooks, Gerrit Blaauw, John Cocke, and Elaine Boehm; Bob Evans reviews plans and asks Amdahl's opinion
  • April: first delivery of Stretch
  • May: Watson makes negative comments about Stretch at news conference while at WJCC; price cuts on 7030 (Stretch) and product withdrawal
  • May 8: memos on "what went wrong with Stretch"; Kolsky writes 14 pages to Emmanuel Piore supporting Stretch
  • May 17: Kolsky writes to A.G. Anderson, Director of San Jose Research lab, that a Stretch follow-on should be built in San Jose; names mentioned as key players were Gene Amdahl, John Cocke, Norm Hardy, and Floyd Johnston
  • May: plans for 8000 series terminated
  • July 11-12: Stretch users group meets Los Alamos: LASL, LRL, AWRE, MITRE, CEIR, Weather Bureau, and IBM; organizations were waiting for the next model
  • July 31: meetings at LRL (and later at LASL) on 10-20 times Stretch; Livermore wanted a new machine in 1963, with word length of at least 48 bits; Vince Learson learns of Livermore bid invitation, but Bob Evans advises against; Learson asks Emmanuel Piore to prepare a ten-year plan [August 31 memo from Learson to Piore?? Pugh, et al., p. 374, n. 10 - but timing is inconsistent]
  • late August: Piore's committee meets and recommends two projects: Project X designated as "ten times Stretch" and an unnamed assignment for Research to study technologies needed for a machine one hundred times Stretch
  • October: Project X transferred to DSD [Fisher, p. 154]; Gene Amdahl leaves Research to work under B. Evans (DSD systems mgr.) on Project X; some engineering veterans of Stretch were initially assigned to Project X
  • November: SPREAD report on unified product line
  • November 28: Piore memo: "The Research organization has accepted the responsibility to bring together background research programs which will provide IBM with the technology necessary to build machines in the range of one hundred times Stretch. The first phase of this program will definitely not be recognizable as a machine development program because it will concentrate on a study of all the basic factors involved in componentry and organization before attempting to specify a machine." [quoted in Pugh, et al., p. 399, n. 112]
  • December: 7040 and 7044 announced
  • December: Amdahl forms committee to study fast circuits for Project X, includes Jim Pomerene and Fred Buelow among others; Pomerene declines Project X manager position and instead transfers to Research; Robert Meade chosen as initial manager

• 1962
  • Stretch follow-ons besides Project X dropped as "too expensive" [H.K.]
  • January: 7094 announced
  • January 11: Amdahl summarizes Project X technology at IBM R&D Board meeting
  • January 23: formal report of "Project X Technology" group
  • involvement of Jim Pomerene? [Pugh, et al., p. 399; but D.S. says Pomerene was never involved with P-Y/ACS work];
  • April 12: Project X production estimated at twenty systems at IBM R&D Board meeting; Board concerned if circuit and memory technologies would be available for targeted announcement in late 1965 and first delivery in 1967; machine cycle time assumed as 25 nsec during mid 1962
  • May: CDC 3600 announced
  • July: reports of CDC contract negotiations with AEC; Fisher describes it as a formal announcement [p. 93]
  • August 16: report by Jim Pomerene of Project Y work (?) [Pugh, et al., p. 399, n. 113; see above]
  • Solomon parallel computer described at FJCC

• 1963
  • 1H63: Project X cost-performance studies
  • August: CDC 6600 announced; 100 nsec cycle time; led to famous "... including the janitor" memo from T.J. Watson, Jr., on August 28
  • August: marketing forecast of 53 machines with power of 10 times a 7090 [Fisher83]
  • August 20: Kolsky's memo to E. Piore, entitled, "Some Comments on the AEC Computer Situation"; memo later used in antitrust suit [see Fishman, p. 121]
  • September 5: (arising from "Jenny Lake meeting" at Jackson Hole, Wyoming) Project X becomes NPL 604 family member
  • September 5: (arising from Jenny Lake meeting) Tom Watson, Jr., directs "Research to undertake the design of a machine of highest attainable performance" [memo of McWhirter to Tucker, Pugh, et al., p. 399, n. 115]
  • September 10: Tom Watson, Jr., Dean McKay, Spike Beitzel, and Harwood Kolsky meet in Los Angeles to discuss supercomputers
  • September 19-20: Machine Organization Committee for Project X meets, includes Amdahl (chair), Chen, Cocke, Hofler, Kolsky, and Meade; cycle time assumed at 60 nsec
  • September 30: memo from G.L. Tucker to E. Piore which says "At Jenny Lake Mr. Watson instructed Research to mount an effort directed toward a scientific processor substantially beyond the 6600 or Project X."
  • November 19: memo from Piore to Watson that CDC 6600 success lay in "singleness of purpose"
  • November: Experimental Computers and Programming, headed by Jack Bertram, becomes home for "an expanded Project Y" [Pugh, et al., p. 399]
  • December: reports of CERN contract for CDC 6600

• 1964
  • Bertram's department studies "ultraspeed computer principles and techniques"; John Cocke and Herb Schorr (and others?) design the initial inst. set; X and A insts. are split into two streams, similar to Stretch; Herb Schorr and Brian Randell (and others?) later work on refining the inst. set
  • April: footnote to S/360 announcement mentions high-performance "upward-compatible" Model 90; McCarter transcripts indicate that the logical organization of the machine was fully specified by this time and that a data flow model had been constructed to test the proposed circuitry [Fisher83, p. 297]
  • June: Livermore requests a Solomon-like machine; Amdahl, Bertram, Cocke, and Kolsky meet to help plan a response; will develop into a proposal by Jim Pomerene for 32-unit Parallel Network Digital Computer (PNDC) for delivery at end of 1966 [see Pugh, et al., p. 408]
  • August: Model 92 announced; J. Galage (engr. mgr.), M. Flynn (CPU design mgr.)
  • August: marketing forecast rises to 70 machines of size and power of Model 92, with expectation of making 24 sales [Fisher83, p. 297]
  • September: first shipment of CDC 6600 (note: Livermore says their first 6600 was introduced in April 1964; Livermore would buy four 6600s)
  • September: Gene Amdahl moves to California with plans to be a visiting faculty member at Stanford for one quarter, given an office in Los Gatos
  • October: Amdahl, Chen, and Conti papers on M92 in FJCC; "cycle time less than 90 nsec"
  • November: Model 91 announced (having less expensive core memory than Model 92 so as to compete more directly with the CDC 6600)
  • December: CDC 6800 announced (will later become the 7600); cycle time estimate of 25 nsec

• 1965
  • January: presentation of PNDC to Livermore
  • Gene Amdahl named an IBM Fellow
  • March 21: Bertram, Cocke, and Kolsky meet to plan for Executive meeting on Project Y
  • March 24: Executive meeting on High Speed Computers; A.K. Watson, J. Haanstra, P. Knaplund, W. Hume, R. Piore, etc.; topics included M92, 6800, David (successor to Solomon), Project Y, etc.
  • April: S/360 Models 60, 62, and 70 withdrawn and replaced by Models 65 and 75
  • May 10: Piore recommends program to reach performance level of "10 times the 6800" with "initial target date of 1968" [Pugh, et al., p. 402, n. 126]
  • May 17: Tom Watson, Jr., "go for broke" memo, deciding "that a suitable response to the CDC challenge required an unfettered development laboratory"; decided with Al Williams on Max Paley as "suitably aggressive project leader" [Pugh, et al., pp. 402-403, memo is n. 127]; "unfettered" => modeled after Cray's success with a small group of top people (at least in Bertram's mind)
  • June 1: Paley calls Kolsky saying that new project (ACS) was approved by executives and was to be located in California
  • June 9: meeting in Oakland: Paley, Bertram, Cocke, Kolsky, and others; Paley proposes a design freeze by end of year, prototype running in mid-1967, and production machines in mid- to late-1968
  • Bob Evans asks Amdahl to take part in ACS
  • June and July: Bertram ill and misses meetings at Corporate Technical Board; Haanstra reports in his place in July; "Paley's project ... intended to build a system with the aid of monolithic circuits, thin-film main memory, and ferrite-core bulk memory ... a heavily pipelined floating-point processor with a 10-nanosecond machine cycle ... second-version system with multiple processors ... completion of prototype system was targeted for the second half of 1967 ... project planned to be underway in a West Coast site within three months ... multiple-head magnetic disks and optimizing compilers" [Pugh, et al., p. 403, n. 128]
  • Summer: Supercomputer lab set up under Max Paley as director and Jack Bertram as design manager; 12 (15?) people move from Watson Lab to the new lab; 30 engineers move from Poughkeepsie and San Jose to the new lab; first facility is on Kifer Road in Sunnyvale [E.S. remembers being in CA on July 4, '65; L.C. remembers apt. hunting in Sept.; R.B. remembers August]
  • July 11: presentation to Los Alamos, which was not well received ("It appeared as if IBM had frozen specs on the next generation computer and were not sincerely interested in additional inputs.")
  • July: all the core-memory versions of Models 91 and 92 are renamed as Model 91; thin-film memory version is named Model 95
  • August 3-7: Arden House Conference, Harriman, New York - full-scale planning for ACS; Project Y presentation includes
    • "air of destiny"
    • multiple decoding and issue with maximum rates being two indexing instructions, two arithmetic instructions, and one branch per cycle
    • FPP (floating point processor) should execute an average of 1.5 instructions per machine cycle
    • "lean machine cycle" - 6 or 7 circuit levels per cycle, 10 nanosecond cycle time
    • branch anticipation using execute-ahead registers
    • technology "Phase II 4Q67: 1 ns ckt ... 30-50 ckts/chip ... liquid cooled ... 16x16 board 27,000 ckts ... FP processor approx. 5 boards"
    • objectives included a 1968-69 delivery with 10x CDC 6800 (i.e., 7600) on sample problems
    • Amdahl discusses having "Two Mountains" to climb and how the system support problem mountain could be more easily scaled with S/360 compatibility
  • branching architecture and lookahead schemes discussed during the year
    • branching architecture worked on by Ed Sussenguth and Herb Schorr
    • (late in the year) dynamic instruction scheduling concepts presented by Lynn Conway
  • August: S/360 Model 67 announced
  • September 8: Kolsky meets with Bertram, et al., at new facility on Kifer Road
  • November 8: Haanstra visits ACS
  • November 17: planning conference on arithmetic, precision, etc.
  • November: Jack Worlton of LASL visits ACS and makes a pitch for vector operations
  • (late) Corporate Technical Board agrees to two year delay in ACS schedules [Pugh, et al., p. 404]
  • (late) NGT availability questioned; Bertram looks to Motorola for improved MECL [Pugh, et al., p. 404]
  • December: Super Machine Task Force ("M91, PNDC, and ACS") formed; includes Amdahl and Kolsky

• 1966
  • January: Don Gibson and D. Pricer circulate technical note on local memory (i.e., cache) for Model 85 project
  • January: Super Machine Task Force meets in Poughkeepsie, Yorktown, and Sunnyvale
  • February: draft of dynamic instruction scheduling paper by Conway, Randell, Rozenberg, and Senzig; scheduling is described in terms of source and destination "sequencing matrices" and a busy vector; the arrangement easily provides for multiple, out-of-order instruction issue; instruction decoding and filling of the matrices stops upon encountering a conditional branch and resumes upon resolution of that branch, and an argument is made for a backwards scan of the matrices in order to give priority to the issue of the conditional branch; scheduling of memory access is also covered; later documents appear to refer to this design as a "contender stack"
  • February 24: Super Machine Task Force meets with Paley and Haanstra
  • June 23: NSA visits ACS; fourteen ACS presenters: Amdahl, Bertram, Cocke, Christopherson, Kolsky, Lloyd, Meagher, Nielson, Pickett, Paley, Reid, Schorr, Steranko, and Sussenguth
  • July 6: final report to Corporate Technical Board from Super Machine Task Force [but see Pugh, et al., p. 405, n. 130, which has March]
  • August 3: meeting on Fortran with Schorr, Allen, Hardy, Kolsky, and others
  • August 23: NSA visits ACS for full presentation
  • September and October: meetings on whether to respond to a Livermore RFP with ACS
  • October 14: Paley memo against any outside commitment of ACS prior to September 1968, responding to pressure from DPD marketing to bid ACS early to Livermore because of competition
  • October: Paley argues for Model 91 derivative over Model 85 in memo to Branscomb
  • November: contract with Computer Sciences Corporation for extended operating system
  • end of year: Supercomputer lab, now known as Advanced Computing Systems, moves to new facilities on Sand Hill Road in Menlo Park

• 1967
  • January 5: first meetings in new facilities
  • January 12-13: IBM Science Advisory Committee meets in Menlo Park, with T.J. Watson, Jr., in brief attendance; full presentation of project; SAC worried that hardware technology would not be ready in time
  • January 13: Kolsky asked by Ralph Palmer (IBM Fellow, corporate director of engineering) over dinner to prepare a report on system and software issues of ACS; Kolsky worked with a small committee, including John Backus and R.J. Creasy, over the next month
  • January: IBM JRD articles on Model 91
  • February 13: Kolsky delivers "Palmer Committee Report"; report weighs questions regarding S/360 compatibility
  • February: Amdahl argues for a S/360-compatible ACS which would use S/360 software [cf. Pugh, et al., p. 405]; result is that he is quarantined by project management and isolated from contact with ACS project members
  • March: ACS delivery estimated in 3Q71, 320K circuits in CPU, 200 MIPS; IOP availability estimated for May 1968
  • April: cycle time estimate now 12.5 ns
  • May: presentation to Palo Alto Scientific Center (ACS described as "up to 7" instructions per cycle)
  • July: I/O computer patent filed (later granted as 3,593,299)
  • major simulation efforts during the year
  • October: first shipment of Model 91, to NASA GSFC; in total, 15 M91s were built (4 were internal to IBM) and 2 M95s; CDC built 94 6600/6700s and 121 other models of the 6000 series [Fisher83]
  • November: ACS presentations to IBM management - multiple decoding and issue with maximum rates being three indexing instructions, three arithmetic instructions, and one branch per cycle
  • December 13: Amdahl argues case for S/360 compatibility to Kolsky

• 1968
  • 1H68: design evaluation of Amdahl's S/360 compatibility proposal
  • January: Model 85 announced
  • January: division patent filed (later granted as 3,591,787)
  • January(?): John Earle is quarantined by project management and assigned to work with Gene Amdahl
  • February: instruction sequence patent filed (later granted as 3,559,183)
  • March 18: Amdahl and Earle argue for a 360-compatible design to Kolsky; proposal known as AEC/360 (for "Amdahl-Earle Computer")
  • March: DPD headquarters asks Kolsky to prepare a position paper regarding S/360 compatibility
  • March: a task force is set up to compare the ACS-1 and AEC/360 designs; the task force members are W.T. Comfort, Carl Conti, Robert Litwiller, and D.M. Powers
  • April 17: Conti prepares performance comparison memo based on task force studies of five benchmarks; AEC/360 comes out better
  • April 26: Kolsky completes "Compatibility Paper for ACS"; states that DPD was not concerned if a few non-S/360 machines were built for scientific purposes, but if the goal were to build and market a large number of them, then DPD would be very concerned
  • May: Bob Evans okays a design shootout between ACS-1 and the Amdahl-Earle design; the review favors Amdahl's proposal and the project becomes ACS-360; consternation within the ranks; Bertram reassigned; several architectural team members go back to Research (including Cocke, Dauber, and Schorr); Amdahl replaces Paley, who leaves and starts a consulting company [date: DeLamarter, p. 220; other info: Pugh, et al., p. 405, and E.S.]
  • June: skip logic patent filed (later granted as 3,577,190)
  • June: series of meetings among Amdahl, Chen, Cocke, and Kolsky
  • July: 90 Series (Models 91 and 95) product withdrawal (press release)
  • 2H68: ACS-360 criticized by East Coast engineers [Pugh, et al., p. 405]; priorities in Poughkeepsie include Model 85
  • summer: Sussenguth investigates two-way multithreading
  • September: ACS presentations at IBM programming symposium (internal) by Herb Schorr, Fran Allen, Jim Beatty, and Charlie Freiman
  • December: CDC files anti-trust suit
  • December: CDC 7600 announced (superpipelined with 3-cycle integer add, 27 nsec cycle time)
  • (month?): CDC starts working on the 8600 [but note 1970 start date; should it be 1968-1972? -- see Gordon Bell's info on 8600]

• 1969
  • January: prepare-to-branch logic patent filed (later granted as 3,577,189)
  • February: Kolsky meets with Chen, et al.
  • "early 1969" Amdahl wants three high-end models in order to return normal profit; top management decides on sticking to the original charge of building one model; Amdahl counters with recommendation that ACS be shut down [DeLamarter, p. 220; Datamation, p. 113]
  • March: first shipment of CDC 7600, to Livermore; although Gordon Bell says Livermore had one starting in January of 1969 (Livermore would buy five 7600s)
  • March: ACS mention in Saul Rosen's Computing Surveys article
  • May: ACS-360 cancellation [DeLamarter, p. 220]
  • "few weeks later" Amdahl argues to Learson, Cary, and Opel for the better price-performance of his high-end designs; they refuse to change the company's marketing policies [Datamation, p. 113; note DeLamarter, p. 221, differs somewhat from this chronology and the one given by Amdahl in March 1973 Computer, p. 37]
  • May: Robelen and Galtieri leave IBM to form Mascor (Multi Access Systems Corp.); Beebe, Buelow, Clements, Tobias, Zasio, and others also leave IBM for Mascor
  • May 27: Harwood Kolsky named IBM Fellow
  • June&July: Datamation articles on ACS cancellation
  • (summer month?): Amdahl bed-ridden with ruptured disc, recovery lasted about nine months [Computer, March 1973, p. 37]
  • August: Model 195 announced
  • August 3-8: task force on vector processing started in Poughkeepsie
  • October: ACS mention in ARPA Technology Forecast
  • December: first shipment of Model 85
  • December: storage transfer, storage control, and interlocking patents filed (later granted as 3,670,307, 3,670,309, and 3,675,217, respectively)

• 1970
  • June: S/370 Models 155 and 165 announced
  • June: storage mapping patent filed (later granted as 3,675,215)
  • August 3: meeting between Amdahl, Kolsky, and Pickett on "Hi-end of line"; last meeting at Menlo Park, later meetings between Kolsky and T.C. Chen and others would take place at San Jose Research Lab
  • September: Amdahl resigns from IBM
  • (month?): Amdahl starts new company
  • December: Amdahl gets $2M from Heizer; Mascor fails and 15 or so Mascor folks join Amdahl
  • December: register renaming on FP load patent filed (later granted as 3,718,912; cf. RS/6000)

• 1971
  • January: first shipment of S/370 Model 155
  • March: first shipment of S/360 Model 195
  • April: first shipment of S/370 Model 165
  • April: Schorr's external paper on ACS at Symposium on Computers and Automata

• 1973
  • introduction of virtual memory on IBM S/370 Models 158 and 168; prevents Amdahl from announcing his completed design

• 1975
  • Amdahl 470 V/6 delivered to NASA

• 1977
  • 3033 announced (design effort was led by Jack Bertram)

• 1990
  • ES/9000 high-end models announced (CPU design similar to ACS-360)

Acknowledgement: I owe extensive thanks to Harwood Kolsky for a timeline he prepared for me in March, 1999.

Navigation within IBM ACS pages:

Back to first ACS page

Next section: Arden House Planning