TIMELINE OF THE IBM STRETCH/HARVEST ERA (1956-1961)* PROLOG 1944: Mark I, the first general purpose digital computer was built at Harvard under the direction of Howard Aiken. The Mark I was used by the Navy for ballistic calculations. 1946: ENIAC (Electronic Numerical Integrator and Computer), with 18,000 vacuum tubes, is dedicated at the University of Pennsylvania. It was 8 by 100 feet and weighed 80 tons. It could do 5,000 additions and 360 multiplications per second. 1946: Eckert-Mauchly Computer Corporation is formed as the Electronic Control Co. to start work on UNIVAC (Universal Automatic Computer). 1948 Fall: IBM produces the "IBM 604", a plugboard programmed calculator using vacuum tubes. In 1949-50 it was turned into the widely-used CPC (Card Programmed Calculator). 1948: IBM builds the Selective Sequence Electronic Calculator (SSEC), a computer with 12,000 tubes. 1950 March: SEAC (Standards Eastern Automatic Computer) is built at the National Bureau of Standards. Was used by Los Alamos scientists. 1951 March: SWAC (Standards Western Automatic Computer) being built at UCLA as a training tool. 1951: The First Joint Computer Conference is held. LARGE ELECTRONIC COMPUTERS GO COMMERCIAL 1951 March: UNIVAC, the first commercial computer was delivered to US Census Bureau by Remington Rand. It used mercury delay lines for memory and reels of magnetic tape for storage. 1951: Whirlwind computer becomes operational at MIT. It was the first real-time computer and was designed by Jay Forrester and Ken Olsen. 1952: IBM TPM (Tape Processing Machine) introduces the first magnetic tape drive for on-line data storage, the 726. It used 8" diameter reels of tape holding about 1 Megabyte, or the contents of about 12,500 punch cards. (IBM was producing 16 Billion punch cards per year in 1952.) 1952 March: The MANIAC I built at Los Alamos becomes operational. 1953 April: IBM installs first stored-program computer, the 701 at Los Alamos. It was a vacuum tube computer with a cathode ray tube (CRT) memory of 2k 36 bit words. A total of 19 were built. The last one was shipped to the U.S. Weather Bureau, Suitland, MD. The 701 established the defacto IBM "price umbrella" which made large computers profitable. 1954: The IBM 650 magnetic drum calculator established itself as the first mass-produced computer, with the company selling 450 in one year 1954 May: The IBM 704 announced as a redesigned 701. It had floating point arithmetic and 4k words of magnetic core storage, later increased to 32k. Gene Amdahl developed the first operating system for the IBM 704. 1954 October: IBM 702, a commercial version of TPM. Followed by IBM 705 in 1956. The 705 III was last of the large vacuum tube machines. THE RACE TO TRANSISTORIZE 1954 Summer: Steve Dunwell joins Ralph Palmer in Poughkeepsie to study "advanced concepts". 1954 October 25: Dunwell and Buchholz issue "Datatron Memo No. 0" saying, "...we must make a giant step..." Several more planning memos followed. 1955 February: The IBM NORC (Naval Ordinance Research Calculator) completed. It had high precision floating decimal and a CRT memory. 1954 December: AEC UCRL Livermore expresses "strong interest" in buying a superspeed transistorized computer. 1955 January 14: VP and engineering director Wallace McDowell holds a meeting to discuss six alternatives for IBM's "next large computer". They were (1) produce NORC, (2) improve NORC, (3) improve 704, (4) modify SAGE, (5) adapt Datatron, or (6) develop a machine 100 times the speed of a 704. 1955 January 20: Ralph Palmer and Cuthbert Hurd get approval to pursue supercomputer talks with potential customers. On January 26 Hurd visits Edward Teller of Livermore. He discusses possibility of a transistorized NORC. 1955 April 20: Cuthbert Hurd makes second proposal to Livermore for a more advanced machine at a later date. It is rejected May 10. (Livermore had already selected Remington Rand.) 1955 June: Sperry and Remington Rand merge to form Sperry-Rand. 1955 August: UNIVAC 1101 operational. 1955 August: The National Security Agency (NSA) gives informal approval to IBM for "Silo," a fast memory development task, and "Plantation,"a computer capabilities study. Term "Stretch" used for the possible 10 megapulse CPU. 1955 September: The Lawrence Radiation Laboratory, Livermore, awards Univac a contract for LARC (Livermore Automatic Research Computer) supercomputer. (The LARC was also offered to Los Alamos later.) 1955 September 20: First IBM presentation at LASL. IBM discussed building a "Ten-Megapulse" computer with 60 bit word, 2 microsecond core memory, fast tapes, I/O Exchange, and maybe a magnetic disk. To take 39 months to build. 1955 November 1: The first in a series of "Stretch Memos" was issued. There were 19 such memos by year-end. Steve Dunwell and Gene Amdahl are major contributors and rivals for Stretch leadership. 1955 December: After learning that Dunwell will be appointed head of Stretch, Gene Amdahl resigns from IBM. (Amdahl will return in 1960 to IBM and work on System/360.) 1956 January 6: Los Alamos Scientific Laboratory (LASL) issues Request for Proposal (RFP) for an advanced computer. IBM's cover letter in its February response includes a design criterion "speed at least 100 times greater than... existing machines". 1956 Jan-Feb: LASL Selection Team visits 4 companies that answered the RFP (3 made written proposals). 1956 April: LASL committee recommends IBM. (Contract discussions take another six months.) 1956 April: First IBM 704 delivered to LASL. OFFICIAL START OF STRETCH PROJECT 1956 Summer: Gerrit Blaauw, Fred Brooks, Werner Buchholz, John Cocke and Jim Pomerene join the Stretch team. Lloyd Hunter provides transistor leadership. 1956 July: In a report Werner Buchholz lists the advantages of a 64-bit word length for Stretch. It also supports NSA's requirement for 8-bit bytes. Werner's term "Byte" first popularized in this memo. 1956: Shipment of the first IBM 305 RAMAC to Zellerbach Paper in San Francisco. This is the dawn of magnetic disk storage. 1956 September 5: First meeting of the Joint IBM-LASL Mathematical Planning Group specified in the Stretch contract. There were a total of 10 meetings. The Los Alamos members were: Bengt Carlson, Harwood Kolsky, Roger Lazarus, David Wood, Robert Frank, and Ed Voorhees. IBM attendees varied but usually included: Gerrit Blaauw, Fred Brooks, Werner Buchholz, and John Cocke. 1956 November: The LASL/IBM contract is formally approved. Delivery is set at 42 months (May 1960) 1957 January: The Stretch project is reassigned from Research to Product Development, allowing Stretch to remain in Poughkeepsie as Research moves to Westchester County 1957 1st quarter: The "3-in-1" was a serious design study in IBM. It envisioned a commercial computer (BASIC) with the either a scientific attachment (SIGMA) or an NSA (Harvest) attachment. The idea was later dropped. 1957 May: The first Harvest manual is published. It summarizes NSA's requirements for computing. NSA proposal acceptance came 12 months later. 1957 August: LASL physicist Harwood Kolsky joins IBM. He and John Cocke design a Stretch simulator to run on a 704 computer. An invaluable tool for Stretch system planners ("architects") and design engineers 1957 November: Stretch project headcount reaches 180. Headcount will grow to approximately 200 in 1958, and 300 in 1959. 1957: Philco S-2000 Transac transistorized computer startled IBM designers. 1957: Control Data Corporation (CDC) is formed by William C. Norris and a group of engineers from Sperry-Rand. 1957: Digital Equipment Corporation (DEC) is founded by Ken Olsen. 1957 November: IBM-built SAGE (Semi-Automatic Ground Environment) links hundreds of radar stations in the United States and Canada in the first large-scale computer communications network. 1957 December: Last meeting of the LASL/IBM Mathematical Planning Group. It declared "design complete". 1958: First IBM 709 delivered. Had separate I/O control called "Data Synchronizer". 1958 May: The first Stretch operations manual is published. This allowed detailed logical design activities to begin. Erich Bloch joins the Stretch project as engineering manager. Jim Pomerene is Harvest's engineering manager. 1958 May: IBM's Harvest proposal is accepted by NSA 1958: Seymour Cray builds the first fully transistorized computer for Control Data Corp., the CDC 1604. STRETCH CONSTRUCTION BEGINS 1959 January: Assembly of the Stretch engineering model begins. Concurrent initiation of the Harvest engineering model and the new transistorized 709 (later called 7090) strains Poughkeepsie lab resources. 1959 May: Stretch LASL/IBM contract renegotiated. The new large parallel Stretch disk was a major plus. 1959 July: IBM's corporate magazine "Business Machines" reports that Stretch will be "about 100 times faster than the most advanced computer working today." 1959: IBM introduces the 1401 commercial system. Over 10,000 units will be delivered during its lifetime. 1959: IBM ships the transistorized computers, the 1620 and 1790. 1959 December: Erich Bloch and other members of the Stretch team present detailed reports at the Eastern Joint Computer Conference (EJCC). Speed comparisons are carefully qualified. 1960 January: CDC 1604 makes its appearance. 1960 April: At IBM's annual meeting for shareholders, CEO Tom Watson, Jr. announces that "IBM will now contract with business firms and government agencies to build Stretch type computers ... [they] will perform more computations per dollar than any other system in the world." 1960: Benjamin Curley develops the first minicomputer, the PDP-1, at Digital Equipment Corporation. 1960 April 28: The Data Processing Division, IBM's sales and marketing arm, places full-page Stretch ads in major newspapers 1960 May: Univac LARC installed at Livermore. 1960 June: IBM 7090 completed. A transistor version of the 709, built of Stretch technology. Developed for BMEWS (Ballistic Missile Early Warning System). It became a great commercial success because of its price, and because it could run existing 704-709 code without reprogramming. 1960 First Half: IBM's Federal Systems Division (formerly Military Products Division) on the downslope of peak SAGE workload, prepares to manufacture and install Stretch computers out of its Kingston, New York facility 1960 December: The AEC orders a Stretch for the Lawrence Radiation Laboratory (LRL), at a price of $13.5 million, delivery in October 1961. The contract calls for floating point operation times in microseconds of.83 (load/store), 1.38 (add/subtract) 2.48 (multiply) and 9.00 (divide), (Plus or minus 10 per cent). 1961 February-March: Tests run by IBM and LRL suggest that Stretch is only 4-5 times the speed of a 7090, against a contract target of 8 times. (In April 1961 IBM management reduced the price from $13,5 million to $7.78 million.) STRETCH SHIPMENTS BEGIN 1961 April 16: The LASL Stretch is shipped from Poughkeepsie to Los Alamos. The first transistorized computer with 64-bit data paths and the first to use eight-bit bytes. It included multiprogramming, memory protection, interrupts, memory interleaving, lookahead, standard 1/0 interface, etc. It remained operational until l971. 1961 May: Ralph Palmer writes a six-page memo putting Stretch into a historical perspective: "IBM [in 1954-55] was behind competition in device and circuit development...the Stretch program was intended to force IBM by commitment into a lead position." 1961: Stretch veterans including Fred Brooks and G.A. Blaauw are developing the "8000 Series". It was later displaced by System/360. 1961 May 9: Tom Watson, Jr. gives talk at the Western Joint Computer Conference in Los Angeles, entitled "Automation and National Power." In his remarks, Watson said, "...And we fell a little short of our optimistic expectations. We're proud of what Stretch is and sorry it didn't end up stretching further." 1961 May: George Monroe, development manager of the IBM 7090, takes over the 7030 project area. Plans for any "Super Stretch" are dropped. 1961 July 11-12: Stretch Users' Group meeting at Los Alamos. People came from LASL, LRL, AWRE, Weather Bureau, MITRE, CEIR, IBM and perhaps others. Returning from the meeting, Harwood Kolsky writes in a trip report, " ...the people at LASL who are actually using Stretch...are almost uniformly happy about the machine. They feel that it is very reliable, ...easy to use, and quite fast." 1961 Summer: Steve Dunwell is transferred to Research in northern Westchester, He begins a study of development methods for computer-aided instruction. He remains there largely unnoticed until 1966. 1961 September: First draft of Werner Buchholz's book (Published 1962). 1961 November: The LRL Stretch is delivered to Livermore, California 1962 January: The Harvest system is delivered to NSA near Washington DC. Included TRACTOR, a large automated tape cartridge library 1962 February: The Atomic Weapons Research Establishment (AWRE) Stretch is delivered to Aldermaston, England. 1962 June/July: The Weather Bureau Stretch is delivered to Washington, D.C. 1962 September/October: The Naval Weapons Lab Stretch is delivered to Dahlgren, Virginia 1962 September: IBM 7094 announced as second generation 7090. 1962 December: The Mitre Corporation Stretch is delivered to Bedford, Massachusetts 1963 August 20: Report on "AEC Computer Situation" by Harwood Kolsky calls for a new Stretch-like project to regain large scientific customers for IBM. 1963 November: The Commissariat a I'Energie Atomique (CEA) Stretch is delivered in France. EPILOG 1964 April 7: IBM announces the System/360, the first family of compatible computers. After a shaky start, System/360 and its successors became IBM's flagship product line for the next 20 years. 1964 April 8: The day after IBM's System/360 announcement, Dunwell wrote to Watson: "The new System/360 is in many respects the image of Stretch." "Also attached is a book edited by Dr. W. Buchholz setting forth the design considerations for Stretch. The book contains sections [written] by several of the principal contributors to System/360 ... " The system features listed included multiprogramming, memory protection, interrupts, memory interleaving, lookahead, 8-bit byte, and standard 1/0 interface. 1964 September: Control Data Corporation introduces the CDC 6600, designed by Seymour Cray. It has 60-bit words and parallel processing. It was the most the powerful computer for several years. 1965 June 1: IBM starts Advanced Computer System (ACS) project in California patterned after the Stretch project. Was cancelled May 1969. 1966 March 15: Steve Dunwell is named an IBM Fellow at IBM's Annual Awards Dinner. Tom Watson Jr., speaking of his 1961 decisions says, "I fear...the great contribution of Stretch to our whole future in IBM got obscured and muddy." 1967 February: The S/360 Model 91 shipped. (About 5 times Stretch performance). IBM reenters the Supercomputer race. 1967 June: Control Data CDC 6800 announced. 1970: IBM/370 series introduced. 1971 February: The S/360 Model 195 announced. 1971 June 21: After 10 years of Stretch service, LASL holds a seminar, "The Historical Importance of Stretch." Stretch computer Dunwell and Buchholz are invited to speak. 1971 May: Brigham Young University (BYU) acquires Mitre's Stretch as "government surplus" 1971 June 29: Tom Watson Jr. retires as chairman of IBM. 1971 November: BYU acquires the LASL Stretch for spare parts 1975: Steve Dunwell retires from IBM and starts own company. 1976 February 27: Harvest is retired at NSA after 14 years 1980 September: BYU retires the Stretch 7030 from its service bureau operation. This is the last of the Stretch 7030 computer systems to retire. --------------------------------------------------------- *Sources include: "Wasn't That a Time - Stretch/Harvest Retrospectives", Edited by Eric Knutsen, September 2002; "IBM's Early Computers", by C.J. Bashe et al, 1986; "History" a web page by Andrew Grygus - Automation Access, 2002; "Third Survey of Digital Computing Systems", US Dept of Commerce, 1961; and personal notes by H.G. Kolsky ---------------------------------------------------------