Memory Size versus Computation Speed for various calculators and computers ffiM1Q90 1M t ow A leD o • lOOK
•• EDVAC lASo SEAC • Whirlwind ...... ---, o ENlAC SWAC /~ " Harvard.' .\ EDSAC o o · o • Mark I • Pilot / 0 \ ACE • • ABC o I I Manchester MKI • • em+l ! oZ3(f1·pt.) ! 1.000 I • oENIAC I lem l i • • Colossus \ I o •\ Bn I (complexV • 100 • 0 • ...... ;' .. -.. -- Comptometer - lme l ' with constants ID U 1.0 ID 100 1K IDK lOOK 1M
GENERATIONS:
ITJ = electronic- vacuum tube Iml = manual m= transistor Ime I = mechanical I]] = integrated circuit lem I = electromechanical m= large scale integrated. circuit
• CONTENTS THE COMPUTER MUSEUM BOARD OF DIRECTORS The Computer Museum is (] non-proHI, Kenneth H. Olsen. Chairman public. charitable foundation dedicated, 10 Digital Equipment Corporation preserving and eJ:hibiling on indus try-wide, broad-based collection of the history olln Charles W. Bachman formation processing. Computer hillory is CuJJjnone Dotabase Systems A CD" a10II to the "-"r interpreted through exhibits. publico:tions, C . Gordon Bell ~n..U- videotapes. lectures, educational p rograms. and other programs. The Museum archives Digital Equipment Corporation 1 Introduction both artifacts and documentation and Gwe n Bell makes the materials available for The Computer Museum 2 Bell TeJephone Laboratories scholarly use. Harvey D. Crogan Modell Complex Calculator The Computer Museum i. open to the public '/exas Insrroments Sunday truough Friday Itom 1:00 to 6:00 pm. 3 Zuae Zl. Z3 There is no charge Jar admission. The Robert Everett Museum's lecture hall and reception The Mitre Corporation 4 ABC. AtanalOff Berry Computer facilities are available lor rent on a mM ASCC (Harvani M Store Personnel: Linda Davidson Merle Insigna Corol Strecker Photo Credits: p.3, from Konrad ZUle; p. 4, top from John Vincent Atanasoll; p. 5. col umn 1. Harvard University, Cruft Laboratory; Th. Comput. , Museum p. 6, from Arthur Burks: p. 7, William M. One Iron Way Ritlase; p. B, Institute lor Advanced Studies, Marlboro, Mas8achuseUs 01752 Princeton University: p. 12, National Bureau 617-467-4036 of Standardll: ppli. 2. 4, 5, 9, 10. bock cover. David BlOmlield 1C 1982ITHE COMPUTER MUSEUM A Companion to the Pioneer Computer Timeline INTRODUCTION Beth Parkhurst carried out the research and compiled the text for This booklet is meant to be an both the exhibit and its companion. exemplary companion: informative, The Timeline has the' specifications attractive and user-friendly. People spelled out for each machine. an who visit the Timeline. a perma overall view of the machine, and nent gallery devoted. to the first. basic descriptive materials. For the one-of-a-kind computers, can use companion, Beth chose quotations this companion as a supplement to that would begin to approximate the exhibit, and those who read it the sentiments of F. C. Williams: "It as the Winter 82183 issue of the Re is fairly easy by reference to such port can gain a sense of the Pioneer records as survive to describe what Computer Timeline. had been achieved. twenty-five years ago. What is difficult is to The concept of developing a recreate the environment of uncer permanent gallery devoted. to the taintyand excitement in which first computers grew out of the those achievements occurred.· Pioneer Computer lecture series ini (F. C. Williams, The Radio and tiated. by the curator, Gordon Bell. Electronic Engineer, 45, 1975. ) The lecture series, archived. on videotape. was led. off on Septem Coordinating the companion. ber 23rd, 1979, by Maurice Wilkes my main concern was to provide a on the EDSAC. Then George Stibitz picture indicative of the Museum's not only talked. about the Bell Tele exhibit. In some cases, I chose phone Relay Computers but repro photographs of the artifacts on dis duced his "Model K ~ adder for the play at the Museum. in others -pe Museum. John Vincent Atanasoff. riod ~ drawings. and in still others the fourth speaker, loaned the the portraits of the machine itself. breadboard Atanasoff-Berry Com And I also insisted on the reference puter for display. With these pieces, page that would indicate where as well as the major Whirlwind primary materials existed on these artifacts and the information gath pioneer computers. While many of ered. at the lectures, the idea of as the original sites of the machines sembling them into an integrated. maintain an exhibit and archives. exhibit emerged.. the Computer Museum is the only place that has an integrated exhibit The exhibit itself, 100 feet long of the Pioneer Computers. on the balcony of the Museum's lecture hall, was designed. and Gordon Bell insisted that this mounted by Jamie Parker and light companion. as well as the timeline, ing consultant Christopher Ripman. put the pioneer computers in their Their concerns were attractiveness historic perspective, and made the and legibility- both for the person comparative table that appears on in the lecture hall and the studious page 14 and the graph on the visitor- as well as flexibility for cover. In compiling the data, we've growth and change. selected from sometimes conflicting information and will keep a run ning corrected. table. The Timeline is not yet com plete. The next Pioneer Computer Lecture, by Captain Grace Hopper on the Harvard Mark I is sched.uled. for April 14. At some time. we will PIONEER integrate video-taped. material O MPUTER into the exhibit; but for now have settled. on keeping it available in the archive. Your suggestions, cor rections, and donations relevant to this exhibit. the lecture series and the archive are encouraged.. - Gwen Bell Director The Compuler MWl8Um Repo,lIWlnler 1983 1 BeU Telephone Laboratories Modell Complex Calculator George Stibitz worked at Bell math course I had learned a little Around that time, the head Labs as a mathematician in the about the binary notation fo r repre of the mathematical engineering 1930s. In his spare time, he experi senting numbers. That notation has group came to Stibitz with a prob· mented with using telephone relays digits with only two values, such as lem. Recent developments in filter for electro-mechanical calculation. zero and one, much as the relay has and transmission line theory were only two 'values': open and closed. overloading the desk calculator -The original notions that led team with complex number work. to the aeries of relay computers had '" It occurred to me that perhaps Could a lorge-scale relay calculator nothing to do with usefulness. I just the two positions of a relay could handle the work? wondered whether it would be pos be used to represent the two values sible to make such simple things of a binary digit. Then perhaps Bell Lobs mode Stibitz's relay as relays do complicated calcula circuits through the contacts of project official with a budget and tions ... several relays might represent the circuit designer. The Model I. first two values of a binary digit. I soon in a series of Bell Lobs relay calcu "I was then a 'mathematical lators and computers. was finished engineer' at the Bell Telephone found out that this was true-two relays could be wired together to in 1939. Technically, the Modell Labs. and as such I was asked to add two binary digits. was not a true computer because look into the magnetic circuits of it was not controlled by a program. the telephone relay. As you know. '"I buUt an adder of the two Rather. it was operated directly a relay is just an electrically-oper. relays I had borrowed. a couple of through a teletype. Although it ated switch that opens and closes dry cells. two flashlight bulbs. and lacked the speed of the electronic one or a dozen electrical circuits. two strips of metal for keys. My computers that were to appear a wife named it the K-mode l. after "'While looking at the relay's few years later. its relays were far our kitchen table. magnetic circuit I naturally noted less liable to failure than vacuum the piles of contacts that could be '"When I took the K-model tubes. closed or opened when the relay to the Labs to show the boys. we The Bell Lobs Modell was the operated. I knew that these COD speculated on the possibility of first demonstration of a lorge-scale tacts could be connected in large building a full-size calculator out of digital machine for complex calcu and complicated. meshes, and when relays. Shortly thereafter the relay lation. so connected. they could do very computer turned serious.- complicated. jobs. So, 1 liberated a -In September 1940, after sev pair of relays from the Labs' junk George R. Stibitz. "Early Com eral months of routine use at the puters and Their Uses,· presented pile and tried out a few circuits. Laboratories. the computer was at Computing and Chili-eating '" Years before in a freshman Society, 1981 demonstrated at a meeting of the American Mathematical Society held at Dartmouth College, in Hanover. New Hampshire ... I gave a short paper on the use and design of the computer after which those attending were invited to transmit problems from a Teletype in McNutt Hall to the computer in New York. Answers returned over the same telegraph connection and were printed out on the Tele type_ '" George Stibitz, "Early Computers," in A History 01 Computing in the 7Wentieth Century, ed. N. Metropolis, J. Howlett, and Gian-Cado Rota, New York, 19110 George Stibitz built this replica of his "K-moder for the Computer Museum. (Gift of George Stibitz, D127.80.) Zuse Zl. Z3 As a civil engineering student "ZUBe describes • . . how his Programs were punched on recycled in 1930s Berlin, and later as an air work was carried out in ignorance motion picture film. craft engineer. Konrad Zuse had to of that of his predecessors. or even spend his time performing ~ big and the contemporary work by Dirks in awful« calculations. Theoretical Germany on magne tic storage sys a d va nces tha t would change civil temB , , , During the war the various engineering from· cut-and-try· to American computer projects were science were starting to appear. but of course subject to strict security were not being applied because of measures; it was only a photograph the volume of computation required that German Military Intelligence in the new approach. Zuse decided had obtained of the Harvard Mark I to build calculating machines to which eventually alerted Zuse to solve these problems automatically. the fact that the Ame ricans had developed some sort of large Bcale tape-controlled compute r. Nothing however prepared him for the post. war release of information about ENIAC. which with its 19.(KIO valves far surpassed anything that he or Schreyer had eve r contemplated attempting to construct." Brian Randell, Th e Origins 01 Digital Computers. 3rd ed .• Berlin. 1982. The designs Zuse began in 1934 led to a series of machines that in cluded the fi rst program-controlled computer. He built a n experimental mechanical computer, the ZI. in the family living room. The Z1. completed in 1938. was followed in 1940 by the Z2, a prototype electro mechanical computer built with second-hand telephone relays. The Z3. a full·scale relay com· puter. was running in 1941. For the first time. the German government aided with funding. This machine had most of the basic features asso· ciated with a conventional com Konrad Zuse examjnes a puter, including memory and a form program lape. of program control. Like Stibitz's electro-mechanical calculator. the Z3 was several orders of magnitude "The work proceeded almost slower than the first electronic com parallel to. but quite independe ntly puters. Its program was external. of. the developments in the United coded on punched film. Two spe States." cial-purpose models, the 51 and 52, were used in aircraft design. Konrad Zuse, · Some Remarks on the History 01 Computing in Ger These first machines were de many,« in A History of Computing stroyed in the war. At the war's end, in the Twentieth Century, ed. N. Zuse learned about the American Metropolis. J. Howlett. and Gian computer ENIAC. and an American Carlo Rota. New York, 1980. observer published a description of a preliminary version otZuse's next relay machine, the Z4. It was not until the 1960s thai an English language account of Zuse's fi rst machines appeared. The Compute. Museum Repo.tIWlnte.l983 3 = ABC IBM ASCC Atanasolf 8eI'ry Computer (Harvard Mark II Beginning in 1935. John Vincent house: use electricity and electron The mM ASCC (Automatic Atanasoff, a physics professor at ics-that meant vacuum tubes in Sequence Controlled. Calculator>' ~ Iowa State College. pioneered digi those days; use base 2. in spite of also known as the Harvard Mark I. tal electronics for calculating. His custom. lor economy; use conden began in the mind of Harvard in students were working with linear sers. but regenerate to avoid structor Howard Aiken. and was partial differential equations. and lapses; compute by direct action. realized by a team representing he experimented with analog. then not by enumeration." Harvard, the U.S. Navy, and IBM. digital calculators to aid in their John Vincent Atanasoli, Pioneer solution. "The desire to economize time Computer Lecture. at The Com and mental effort in arithmetical • "I tried again and again to sort puter Museum. November 11, 1980 computation. and to eliminate hu- these concepts out. Nothing seemed man liability to error. is probably to work. After months 01 work and. as old as the science 01 arithmetic study J went to the office again one itself . .. evening but it looked 08 if nothing would happen. I was extremely "The intensive development distraught. Then I got in my auto of mathematical and physical sci· mobile and started. to drive. I drove ences in recent years has included hard &0 I would have to give my the definition of many new and use attention to driving and I wouldn't ful functions. nearly all of which have to worry about my problem• . are defined by infinite series or other infinite processes. Most of "When I finally came to earth these are tabulated. inadequately I was ero.Bing the Mississippi and their application to scientific River. 189 miles from my desk. You problems is retarded thereby. couldn't get a drink in Iowa in those Professor A tanaself lecturing to days. but 1 was crossing into Illi "The increased accuracy 01 students at Iow a State University nois. 1 looked ahead and there was physical measurement has made in the late 193Os. o light and. 01 course. it was a tav necessary more accurate com ero. I went in and got a drink. and. putation. Many of the most recent Atanasoft and graduate stu then I noticed that my mind was scientific developments are based dent Clifford Berry built a prototype very clear and sharp_ I knew what on nonlinear effects. All too often ~ ABC (Atanasoff-Berry Computer) in I wanted. to think about and I went the diHerential equations designed 1939. and a full-scale model in 1942. right to work on it and worked lor to represent these physical phe Like the Bell Labs Model I, the ABC nomena may be solved only by 3 hours. and then got in my car and. was not a computer in the modern drove slowly back to Ames. numerical integration. This method. sense, since it lacked program con involves an enormous amount of '"I had made four decisions trol and was not general purpose. in that evening at the Dlinois road computational labor. Many of the The ABC was the first of sev computational difficulties with eral proposals to use electronics for which the physical and mathe- Atanaseff built this simple model of calculation or logic in the decade matical sciences are faced can :he ABC to demonstrate his concepts after Atanaseff began investiga be removed by the use of suitable of digital computation. The n umber tions in 1935. Other projects and automatic calculating machine ry. stored in one of the capaCitor drums proposals included those of Bush is added to or subtracted from the and Crawford both at M.I.T.; Zuse "The development of numerical number stored in the other drum. and Schreier in Berlin; the British analysis. including the techniques (On loon from J. V. Atanaself. X12.80.J foreign office; Rajchman at R.C.A. of numerical differentiation and in The makers of the ENIAC. the first tegration. and methods for solving electronic computer, were familiar ordinary and. partial differential with Atanasoff's and Rajchmon's equations have reduced.. in effect. work. The degree to w hich the ABC the processes of mathematical influenced. the ENIAC design is still analysis to selected. sequences of being debated. by participants and the five fundamental operations of historians. arithmetic: addition. subtraction. multiplication. division. and refer- ence to tables of previously com- puted results. The automatic se· quence controlled calculator was designed. to cony out any selected. sequence of these operations under , completely automatic control." Howard Aiken and Grace Hopper 1946 Electrical Engineering 4 Th. Comp!.ll. , Mu.eum ReportlWlnte, 1983 Colossus Inspired by Charles Babbage's "If you hated Hitler enough. United States after the war. The """""'nineteenth-century "Analytical you would light on against learful Bletchley Park effort. however. did Engine." the Harvard Mark I was odds. You considered not just tum out a number of scientists ex mostly mechanical. Counter wheels the small probability 01 success. perienced in electronics and logic. were electrcrmechanical. and but the large payoff if you were F. C. Williams, head of the postwar connections between units were successful... Manchester University computer electrical. An external program project. remembered help he re 1. J. Good, "Pioneering Work on ceived from two Bletchley alumni punched on tape controlled opera Computers at Bletchley," in A His tion; conditional branches were not tory of Computing in the'I\.ventieth who were o1so familiar with possible when the machine was Century. ed. N. Metropolis, American computer projects: "Tom first in operation. The machine was J. Howlett, and Cian-Carlo Rota, Kilburn and I knew nothing about largely built of standard IBM equip New York, 1980. computers, but a lot about circuits. ment. It was completed at IBM in Prolessor Newman and Mr, A. M, 1943, and moved to Harvard in 1944. Th.ring in the Mathematics Depart ment knew a lot about computers The Harvard Mark I's contribu and substantially nothing about tion was not in its technology-the electronics. They took us by the electronic ENlAC, which would sur hond and explained how numbers pass the Harvard Mark I's speed by could live in houses with addresses several orders of magnitude, was and how if they did they could be under construction when the Mark I kept track 01 during a calculation_" was being dedicated. F. C. Williams, "Early Computers at Manchester University." Radio and Electronic Engineer, 1975 Pulley from a Colossus tope drive. (Gift of Toby Harper, X4 9.82.) This spirit motivated the British Foreign Office's cryptanalytic effort at Bletchley Park. German forces relied on variants of the ENIGMA machine for enciphering in World War II. The simplest version of the Intercepted German messages were ENIGMA hod 9 x 1020 initial settings, punched on paper tape and read so breaking the cipher was on awe into the Colossus photoelectrically somely complex process. The Brit ish built a series of machines to "The value 01 the work 1 am decipher intercepted. German sure to engineers like myself and messages. The culmination of the possibly to mathematicians like series was the Colossus line, elec Alan Th.ring. was that we acquired tronic machines with many of the a new understanding of and fa features of the computer, including miliarity with logical switching electronic circuits for Boolean logic. Re-assembling the machine at and processing becau.se of the counting, and binary arithmetic; enhanced possibilities brought Harvard, March 10. 1944. automatic operation, with logic about by electronic technologies functions set with plugs and which we ourselves developed "It is important because it was switches, or conditionally selected Thus when stored program com the lirst large scale digital calcula by electro-mechanical relays; and puters became known to us we tor ever built and also because it electronic registers changeable were able to go right ahead with stimulated the imagination and in by an automatically controlled their development." terest of the world and thus gave sequence of operations. impetus to the desire lor more and T. H. Flowers. letter to Brian Ran The first official release of in ,-..., better computing machines." dell. February 15. 1972; quoted in B. formation on the Colossus was not Randell, "The Colossus." in A His G. Truman Hunter, "Modem Com until1975. Because of this secrecy. tory of Computing in the'I\.ventieth puting Machines,' Journal 01 the the Colossus did not directly influ Century, ed. N. Metropolis, Franklin Institute, 1952. ence the computer projects which J. Howlett, and Gian-Carlo Rota. flourished in England and the New York, 1980. The Computer Muaeum Reporltwlnle. 1963 5 = ENIIlC Each of these earlier machines The ENlAC was moved to the cradle of a whole lot of computers: ,-.... had some of the features of the Army's Aberdeen Proving Ground the EDVAC. ORDVAC. and a bunch . 1 electronic computer. In the ENIAC. after a year of operation at the of others. But even after they were these features-electronic. high Moore School. R. F. Clippinger, a delivered. the ENIAC continued to speed operation, general-purpose mathematician who devised some work for about ten years. There was capability. crnd program control of the first applications at Aber a period when the ENIAC was the were combined. It is usually re deen, recalled.: only computer working. A lot of oth- garded as the first true electronic ers were on the drawing boards or "I had a couple of girls with computer. The major difference be in the mill being engineered. but tween the ENIAC and later comput desk calculators working out the not working." test case that I would use to find ers was that it was programmed by out if I was getting the right an R. F. Clippinger. gallery talk at plugs and switches. rather than the Computer Museum, September running a stored program. swers from the ENIAC. It took them two man-years to do one solution. 26. 1982 The ENIAC, funded. by the We put it on the ENIAC, and the Army Ballistics Research Labora ENIAC ran off a case very hour . . . tory at the University of Pennsyl vania's Moore School. used elec "You have to realize that the tronics on an unprecedented scale. Aberdeen Proving Ground was the Its IB,(X)) vacuum tubes belied the criticism that. given the failure rate of vacuum tubes. one or more tubes The ENIAC team, headed by]. Pres would fail before a computation per Eckert and Joh n Mauchly, in was completed. The success of cluded a dozen engineers and pro electronics for large-scale computa grammers. Designer Arthur Burks tion inspired a number of postwar looks on as a program is set up on computer projects. the ENIAC with plugs and switches. 6 The Computer Museum Report/Winter ]993 . EDVAC The EDVAC was the successor The mercury delay line mem· was widely circulated. in draft. to the ENlAC. While the ENIAC was ory. borrowed. from radar to utilize The Moore School's 1946 summer being built. its designers realized as computer memory, was the key lecture series on the EDVAC design the potential of the stored program. device that made the stored. pro also helped publicize the idea of They began designing a new com gram practical. The ENlAC had the stored. program computer. puter. a nd were soon joined by dis only twenty words high·speed The EDVAC, while still in its de· tinguished mathematician John von memory capacity. using expensive sign stage, directly or indirectly ,--... Neumann. vacuum tubes- far too few to store influenced all postwar computer programs and data. In contrast, projects. The question MWho invented each delay line could hold hun· the program?" has been answered dred.s of words. with bits circulat· many ways. It cannot be attributed ing as ultrasonic pulses in a col to any single person, but seems to umn of mercury. When each bit have arisen in the course of conver reached the end of the column. it sations among ENIAC project mem was converted to an electrical sig· bers; other researchers may also nal. where it was cleaned up and have independently conceived the could be read. idea. Arthur Burks. who worked on The EDVACs theoretical design and the ENIAC. beginning of the EDSAC. Von Neumann's write-up of construction stage lasted from 1944 and with lohn von Neumann on the the EDVAC group's discussions to 1951. lAS computer. made this assess ment of the process of making the stored program practical. "There were two main steps. Pres and John (Eckert and Mauchly. of ENlAC) invented the circulating mercury delay line store. with enough capacity to store program information a s well as data . Von Neumann created. the first modern order code and worked. out the log. ical design of an electronic com· puter to execute it." r- Arthur W. Burks, -From ENlAC to the Stored· Program Computer,· in A History of Computing in the 7Wentieth Century, ed. N. Metropolis, J. Howlett, and Gian·Carlo Rota, New York, 1980. The Computel MUMUm Report/WInter 198:1 7 - lAS Computer John von Neumann left the IDVAC project to return to the Insti tute for Advanced Study, bringing with him Arthur Burks ond Herman Goldstine. The three elaborated. stored program computer design with the draft of "Preliminary Dis cussions of the Logical Design of an Electronic Computing Instrument." The lAS Computer introduced. cially since these correspondents The lAS computer. asynchronous operation. For fast were mostly well established memory. it used the Williams tube, and supported by facilities and a CRT memory developed at Man resources wholly lacking chez nous. chester University. The Williams We anticipated that any mistakes tube was used in serial mode at we might make in sending out Manchester; the lAS Computer piecewise the fruits of our efforts it was first to use in parallel. would thereby be exposed. to possi One of the lAS Computer's bly hostile or competitive criticism. most significant contributions was leaving us no place to hide. but as a pattern for other computer proj in fact problems oj this sort never ects. Julian Bigelow, who was the arose. and communication with all computer's chief designer, recounts: people at these laboratories was "Another feature of the arrange entirely friendly and stimulating," ment for financial support [by military agencies and the Atomic Julian Bigelow, 'Computer Energy Commission] provided that. Development at I.A.S. Princeton,' as sections of the computer were in A History 01 Computing successfully developed. working in the 'IWenlieth Century. drawings would be sent out by our ed. N. Metropolis. J. Howlett, and Gian-Carlo Rota. New York, 1980. engineering group to five other de· velopment centers supported by similar government contracts. no • tably to Los Alamos Laboratory. the University of Illinois. Oak Ridge National Laboratory. Argonne National Laboratory. and the Rand Corporation, For the first year or so this requirement that what we pro· duced was in effect going to be duplicated at five distinguished laboratories elsewhere added to the anxieties of the lAS team, espe- 8 Th. Comp!.l!.' M.... l,lm ReporllWin! •• 1983 EDSAC -r- -The EDSAC i8 baaed on prin ·We realized that building ciple. first enunciated in an unpub the machine was only the start of lished report ... in which ideas lor the project: that there was a great a machine known 08 the EDVAC deal to be learnt about writing were I18t out." programs. about how to use the Maurice Wilkes "Programme machine for numerical analysis. Design for a Higb Speed Aulo numerical calculation, and all the matic Calculating Machine," Jour rest of it .. . As BOOn as we started nal of Scientific Instruments 1949 . programming. we found to our sur ... prise that it wasn't as easy to get programs right as we had thought. ~. By 1949. a number of computers were underway. Maurice Wilkes. Debugging had to be diacovered. Director of Computation at Cam I can remember the exact instant bridge University, was the first to when I realized that a large part of complete a machine with the first my life from then on was going to Valves (the En glish equivalent of program running on May 6th of be spent in finding mistakes in my vacuum tubes) on the EDSAC mem that year. Maurice Wilkes started own programs.· ory driver. Maurice JA.1lkes Is on the the project on his return from the Maurice Wilkes, Pioneer back cover holding the memory 1946 Moore School lectures on the Computer Lecture, The Computer driver's wiring. (On loan from the EDVAC design. Returning to Cam Museum, September 21. 1979 Science M useum, London .) bridge University. he set up the Computation Laboratory and started. work on a stored program computer. Wilkes used. existing technologies to get a machine up and running. His decision on mem ory technology was characteristic of this design philosophy: "We used ~ the mercury delay-line because it was really the only thing you could count on at the time.· Maurice Wilkes. gallery talk, at The Computer Museum, July 7, 1982 ' EDSAC memory delay lines plugged into this tank cover. (On loan from the Science M useum, London.) r Th. Compute. Museum ReportlWinter 1983 9 - Manchester University Mark I Graduate student Dai Edwards. A \rVilliams tube set in the machine con be seen in the foreground. Computer work began at Manchester University in late 1946. F. C. Williams and Thomas Kilburn's first project was to build a new kind of memory, one thai was large enough to store programs and data, but faster than the mercury delay line. Several investigators. most notably Jan Rajchman of RCA. had been working on cathode-ray tube memory. Williams and Kilburn solved a major drawback to the CRT. i.e., that the charged spots that represented bits only stayed on the screen for a few instants before dissipating. -Looking back. it is amazing how long it took to realize the fact that if one can read a rec ord once. then that is entirely suf ficient lor storage. provided that what is read can be immediately rewritten in its Original position." F. C. Williams and T. Kilburn, paper presented at Manchester University Computer Inaugwal Conference. 1951 The Manchester group built an on two levels. The electronic store Williams tube memory was experimental prototype to test the was in the magnetism room and the borrowed by several computers of Williams tube. The "baby machine" magnetic store in the room above. the day, including the lAS Com ran its first program in June 1948. 1iansiers between the stores were puter. Julian Bigelow, head of The machine was expanded in sev achieved by setting switches. then engineering design for the lAS eral stages, and the full-scale com running to the bottom of the stairs project, recalled his visit to see puter was complete in late 1949. and shouting, 'We are ready to re the Manchester Computer in ils Williams described its not-quite ceive track 17 on tube I: The pro early slale: automatic operation: cess was repeated for tube 2 and "My visit to Manchester was "The two-level store (fast the machine set working. When the a delightful experience: F. C . Williams tube and slow magnetic machine wished to disgorge infor Williams was a true example of mation. it stopped and the reverse drum] 1 have referred to was indeed the British ' string and sealing wax' process was initiated." inventive genius. who had built a F. C. Williams, "Early Computers primitive electronic computer from at Manchester University. Radio surplus World War U radar parts and Electronic Engineer. 1975 strictly on his own inspiration- in the middle of which were two cathode-ray tubes staring digits in serial access mode-the 'Williams memory.' I can remember him ex plaining it to me, when there was a flash and a puff of smoke and everything went dead, but Wil liams was unperturbed, turned off the power. and with a handy solder ing iron. replaced a few dangling wires and resistors so that every thing was working again in a few minutes." Julian Bigelow, "Computer Development at LA.S. Princeton." in A History of Computing Williams tube memory. (Gift of fhl:: in the 1Wentieth Century, ed. Department of Computer Science, N. Metropolis, J. Howlett, and University of Manchester.) Gian·Carlo Rota. New York, 1980 10 The Computer Museum ReponlWinter 1983 Pilot ACE ,...... , After the war, Britain's National Thring designed. several ver Physical Laboratory began a com sions of a computer, but left the NPL puter project. Alan Thring. who had in 194Z An NPL team directed. by 1. written a paper on machine intelli H. Wilkinson built a pilot version of gence in 1936 and participated. in the ACE. which embodied Thring's the Bletchley Park cryptoanaiytic ef highly original design philosophy. fort, was the central figure in the Thring summed it up in a 1947 can· early days of the NPL project. In the ference discussion: "We are trying words of the NFL's director, "About to make greater use 01 the facilities twelve years ago. a young Cam available in the machine to do all bridge mathe matician. by name kinds of diHerent thinga simply by Turing. wrote a paper in which he programming rather than by the worked out by strict logical princi addition of extra apparatua." ples how far a machine CQuid be imagined which would imitate Discussion of *Transfer Between External and Internal Memory· processes of thought. It was an by C. Bradford Sheppard. Pro· idealized machine he was consid ceedings of a Symposium on ering. and at that time it looked 08 Large.&ale Digital Calculat if it could never possibly be made. ing Machinery, Cambridge, From Alan Turing's ACE notebook. But the great developments in Mass., 1947. Pin the ACE. we intend to represent wireless and electronic valves dur all numbers in the binary system. ing the war have altered the pic Every number may be represented ture. Consequently. 'lUring. who is in the binary system by a sequence now on our slaH. is showing us how of digits each of which is either a to make his idea come true.· zero or a one, and this provides us with a particularly simple method of Sir Charles Darwin. representing a number electrically." SBC broadcast. 1946 J. H. Wilkinson, Progress Report on the Automatic Computing En gine. Mathematics Division, Na· tional PhySical Laboratory. 1948. , , J + • ...... " ••. ...', ,'. " ."" ."" .' .' .... ' ", ..• ...... " ...... , ...... " . "'N,' "'. .'. , .. ... ,,,...... , ...... , ...... '" .''', .. ~., .''', '.' .' , ..... " ...... , ...... " ...... ,." -..-- ... -...... ~ ,... . •• " ...... , .. ,~ _ ._...... , __ .. "" ..... n •• "N ...... ~~" ...... " .' ... " .. ,~ .'_ '.~...... ' ...... ~ .. .. •" ...... -••• ..,.. . H ..... ' ...... _ ...... _ ...... "u _._. • "" ,," "" "" .m " . ~ _ ...... , ..... _._ ..,, _ .... "" .. " ...... ""' ...... -#.,,..... ", .... -,-,,,. .... "" '''' """" "" '''' ----- ...... ,.' " ", ... ~ ..... "' ." ...... "' ...... ~ ...... -.---,...... ' ...... _-- ...... " ~• ..• "'. 'N' ,... •. N"' .... "" .. " ._ ...... W' ...... _, ,_,_ ...... _ .... __ '- .. " .. " .". " " "" .. " ..... ,,' .... - ,,, ... ,, ----_ .... - - ....<_ _--"" .." "" "" .,,, ., ...." I .., .... _ . __ -_... _--"" .." "" -.-.--.-- ...... ~ ------__ ...... " " ,," , .. , .'''' ...... " ,," "" , .. , II... 11, "" .. " .... "" "" "" ... , III. _ ---_- __ - "" ""...... "" "II. ____----- "" .... --_.' .. :::: :::: ::::-.i , .... _. ,... _...... • _II" ...... " .. " .." "" -'1' ''' "" "" ,,, . .." ,.. ,I".. "" .,It "It 'III "" """" .. ,..l._ "".",_._--. ----'"'1''' "" .,., ,10-0 .... ._- I .I The Computer Museum Repor t/Winter 1983 11 - National Bureau 01 SlandCll'ds SEAC and SWAC Before any of the stored pro objective was to be a laboratory for established. two standard debug gram computers had been com testing components and systems. ging techniques. After about two pleted, the National Bureau of since the Bureau of Standards hours a day of preventive mainte Standards decided to procure two might be called. on to set standards nance. we would start 6: test pro computers for its own use. After relotiog to computers. gram running. Then we applied the reviewing university projects and SWAC (Standards Western 'stir with a wooden spoon' tech proposals from nascent computer Automatic Computer) was built at nique. which consisted of taking companies, Standards decided the Institute for Numerical Analysis something like a wooden spoon to build their own machines. in Los Angeles. Its main objective and going around the computer. was to be finished as soon as tapping everything you could see. If the test program stopped, you possible. using as much already developed. technology as possible. had found something. When that Project leader Harry Huskey wrote. test was finally passed. we applied "The plan was to build a computer the Bureau of Standards' 'standard jump.' We were in a building with with the minimum of circuit devel opment. Thus. the circuits in the wooden Doors that were not diHi arithmetic unit were derived from cult to shake, so the standard jump Whirlwind circuits, and the devel consisted of jumping up in the air opment of the memory circuits de about 15 cm and coming down on pended heavily on the published the floor as hard as possible. If that work of F. C. Williams of Man test was passed, the machine was chester University." ready to tackle a computational program-and even more interest Harry D. Huskey. "The National ing bugs would show up." Bureau of Standards Western Automatic Computer (SWAC)," Ralph J. Slutz. "Memories of the in A History of Computing Bureau 01 Standards' SEAC." in A in the 'IWentieth Century. History of Computing in the TWen· ed. N. Metropolis. J. Howlett. and tieth Century, ed. N. Metropolis. J. Gian-Carlo Rota. New York. 1980 Howlett. and Gian-Carlo Rota. New York. 1980 SEAC was the first computer SEAC was the first of stored. to use all-diode logic. pointing the program computer to be completed way for the solid-state computers in the United States. followed of later years. Diodes were much shortly by SWAC. With the first more reliable than vacuum tubes. SEAC console. English computers. the Standards The SEAC. however. required a computers reassured workers on The SEAC (Standards Eastern good deal of maintenance. like all other contemporary computer proj Automatic Computer), built in computers of the day: -We actually ects of their feasibility. Washington. had two aims. One had much more trouble from bad was to be operational as soon as solder joints than we ever had from possible to run programs for the vacuum tubes, diodes, or delay Bureau of Standards. The second lines. I can well remember that we SWAC block diagram. 12 Th. CompUter Mu ..um ReportIWlnter 1983 Whirlwind In 1944. the Massachusetts An elaborate system of mar 25 microseconds) thus increasing f""". Institute of Technology contracted ginal checking identified hardware the speed of computer operation." with the Navy to build a universal problems before they affected com M.I.T. Project Whir.lwind, aircraft flight simulator/trainer. puta1ional accuracy. Summary Report #35, 1953, p. 33. Jay Forrester of the M.I.1. Servo Institute Archives and Special mechanisms Lab became director At the some time, new military Collections, M.I.T. Libraries, of the project. By 1945. the original applications which demanded Cambridge, MA. conception of an analog machine higher-than-ever reliability were was dropped, and the Navy ap- emerging. The Cold War was at its Whirlwind was thus the first proved construction of a digital height, and the U.S. military was full-scale computer to run on core computer in 1946. A general-pur on guard against atomic attack. memory, the mainstay of primary pose computer could take care of Whirlwind, funded by the Office memories until the 1970s. not only flight simulation calcula· of Naval Research and then by the tions, but a variety of other sden· Air Force, was part of the defense tific and engineering applications. network; the production version Whirlwind was completed in of the Whirlwind II design, named stages; the entire central machine ANIFSQ-7, was to become part of was working in 1951. the SAGE System. Project members, dissatisfied with CRT memory per The most important legacy of formance, researched a substitute. the flight-simulator concept was Whirlwind's real-time design. To Several researchers in the late allow the instantaneous response 1940s, including Jay Forrester, con needed for flight simulation, Whirl ceived the idea of using magnetic wind originally used its own ver cores for computer memory. Wil sion of cathode-ray tube memory, at liam Papian of Project Whirlwind that time the fastest available type dted one of these efforts, Harvard's of memory. It was also, in the words "Static Magnetic Delay Line," in an of a 1952 project summary report, internal memo. Core memory was "the most important factor affect installed on Whirlwind in the sum ing reliabUity of the Whirlwind I mer of 1953. "Magnetic-Core Storage "'"" system." has two big advantages: (1) greater reliability with a conseqflent reduc M.l.T. Project Whirlwind. Summary Report #31. 1952, p. 6. tion in maintenance time devoted Institute, Archives and Special to storage; (2) shorter access time Collections, M.I.T. Libraries, (core access time is 9 microseconds: Cambridge, MA. tube access time is approximately 'he Comp.lI... MUMl.UD R.portIW\nterl983 13 - The Pioneer Computers Comparative Statistics Start up Completion Word. length Bell Labll Model I 1939 10/39 4 function, 8 digits George Stibitz at complex Bell Telephone arithmetic Laboratories calculator Zuse Z3 1939 1941 punched film 22 bits, Konrad Zuse fit. pt. ABC 12 /37 12 /39 fixed.. 50 bits John Vincent Atanasoff prototype equation and Clifford Berry at 1942 solver Iowa State University IBMASCC 1937 8/44 punched tape. 23 digits Harvard Mark I function table. also double plugboard precision Colossus {Mark I &: m 1943 12/43 (I) telephone 5 bit Bletchley Park 5144 (ll) plugboard (I), characters switches (II) ENlAC 1943 2146 plugboard. 10 digits Moore School. switches University of Pennsylvania EDVAC 1144 ISSI stored 44 Moore School. program University of Pennsylvania computer lAS Computet 6/46 7/51 40 Institute for Advanced. Study, Princeton University EDSAC 10/46 5/49 Maurice Wilkes at Cambridge University MANCHESTER U. MARK I 1947 6/48 40 Manchester University prototype 7/49 PU.OTACE 10/48 5150 32 National Physical Laboratory, Teddington, England SEAC 6/48 5/50 45 National Bureau of Standards SWAC 1149 7/50 41 National Bureau of Standards Institute for Numerical Analysis Whirlwind 1945 ISSI 16 Servomechanisms Laboratory, MIT ------~ In many cases, the materia l in this table was compiled from data sheets filled out by a person who worked on the machine. In the literature conflicting published data was found. The Museum requests corrections to this table and will keep the most aCCUIate updated version on file. 14 The Compute. Museum Repo.tIWinte.l9B3 The Computer Mu.seum ReportlWinter 1983 15 Additional s-roe MaterIal PrimeR}' .aurce boob with ex Bell t.l.ph_ Laboratort•• Mod.! I National MUMUm at American History. in the SEAC.· Malllell'lOliro/l:lhles and cellelll bibUogrophl••. guldiDg the Goorge R. Shbilz, videotape of lec1ure a\ Smith!lOfUan ImtitUtlon. Washington. Other Aids to Computalion S. 1951 . . ~ Na1Mr \0 gf'MII DWDbe ... o! prI.mcuy Thl> Computer MUiI&um. 1900. D.C. Ernest f. Ainsworth. 'Operational Expefl· cmd MOOadcuy aoun:•• : Gvolve Robert S!ibltz pope". Dortmooth Donald Ead~ . 'EDVAC Drum Memory &nee with SEAG.' Proceedings oIlhe College Ubrary. Phase Sywtem 01 MagnetIC RecordIng.' Joint AIE£.IR£·ACM Cbmpllte, Con/ft" ArchIves. BeIl1i>1ophone Loboratortes. £/ecT.irol £ngme"';ng n. 1953. ene.. New Ya"-':. December 10./2, 1952. C. Gordon BeD and Allen N__ U, S. GreenwaJd. 'SEAC Input-Output G.R. Stiboll. 'Colcukning With ltlephone s. E. Gh.d. 'The ElectrorUc: Di!ICJete Comptl"'" S!ru<;ture.; Reodings and \briobIe Compulef. ' £ktct.irol E'nginew. System.' PtoCMdings oI t'" Equ\pm9nt. • Po:lper presen1ed a\ McJthe.. Jomt Example.. New YOrk. 1971. AJ££·IR£·ACM Computer Con/erenoe. maI;,;,;,J Aalctia!ion America IT-'U'IQ'. mgn. 1953. B. V. Bowden. Editor, foster than of New "!btl:. Deeembe, 10·12. 1952. Honover. N.H.• 1940. lAS Camputer Though!. A Symposium on [);glrol Ruth C. Haueter• . AUXIliary Equipment Computing M achina. New York. 1966. ZIlH n , Z3 'The lAS Computer ~ on exhibt at the to stAC Input·Output· Proceedings 01 N. MeltCpolil. J. Howloll. gntl A ",plica altha Z3 !son exhibit gt the National Museum 01 American HIStory. Ihe join! AI££·IR£·ACM Comput., Con· Smilhllonkln m.tltutlon. Gian·CarIo Rota. Ed.il0r:0. A Hllrory DeutlCh.. Museum. Munx:::h. Washing1on. fervnce. New York. December 10·12. 01 Compurmg in rhlt 1Wemlelh Cenlu~ D.C. Konrad Zuae. v\dooIape 01 lecture at The 1952. New Vorl:, 1900. Campul... MUNUm, 198\. M01hernatiCI and Natural Scienc:w 10""", L Pike. 'lnptl!·Outptlt Devices Ubrary; m.lltUI8Iof Advanced Study. Brion RandeU. rdilOf. 1lNt Or/g'ln&" 01 K. Z.... . Caku/afOl fat 1echnical and U-.,d with SEAC: Proceedings 01 !he Digiral A;:pe. .. P'rlncelOn. N.I. Comp.Jlers. Sel«ffd Sotmttl>e Colculatlon.f DessgoMd Acxon;I· Jain! AI££.IR£·ACM Compurer Conte,' Thlrd Edition, Bertin. 1982. ArchiVflS, DlvWon 01 Mathematics. New "!btl:, 1(}'12. 1952. ing /0 a ~I Plan. Distributed enee. December by the O!fle» of tho Publiootlon Board, NaiJaoal Mu.um 01 American Hlsto.y. Sodrwy GreenwaJd. R. C. Haueler. and Du!xtrtmenl 01 Commerce, 'IobshinglOn. SmithlJonlon InstitutJan. 'Mlshingk>n. S.N. Alaxande,. 'StAC. ' Proceedmgs D.C. (n.d.). D.C. 01 rlat IR£ 41. 19S3. "C "'SAC NlSSWAC A Ilimpllfied model 01 the Alaooso/l· Pans at the EDSAC are on exhibit at Pa". ollhe SWAC are on exhibit otthe The Computer Museum. &trry Computer buill by J. V. Ata ~ Natlonal Bureau at Standarda Museum bon exhibll at The Compute. M UlM\lm. M. V. Wilkh. Yidealapoo oIledure 01 The a nd the Museum 01 50enee and I. V. Alanad. vidootape 01 t.dure 01 The Computer MUMUm. 1979. Industry. Los Angelel. Compute. Muwum. 1900. 'The EDSAC fUm.' Produced by Cam· Ubrmy Division. NatiJogl Bureau 01 Ardliva. Division 01 MathGIrI 16 The Computer MIlHWII Jt.portIWintfi 1983 = rORPORATE FOUJll)ERS Donating $2.500 or more: Benton and Bowles Bolt. Beranek and Newman SYMPOSIUM ON ARCIDVING Robert Cipriani Associates THE IDSTORY OF COMPUTING Clint Clemens ComputerWorld Control Data Corporation Coopers and Lybrand. Boston On May 5 and 6. 1983. the Museum will host a Data General symposium that addresses questions on archiv Digital Equipment Corporation ing the history of information processing. The General Systems Group, Inc. purpose is to share information about activities Intel Corporation MITRE Corporation planned and underway at the museum. insti Richard Reno tutes. libraries. government agencies. corpora Schlumberger Foundation tions, universities, and individuals. Tobin Vending Service AUendance will be limited in order to gain a broad cross-section of information. Anyone in NEW IJO)IVIDUIlL terested should contact Chris Rudomin at The FOUJll)ERS: Computer Museum. (617) 467-7570. The Museum welcomes the follow ing who join the first 86 individual ~ounders by donating $250 or more. o Jarlan and Lois Anderson J. Weldon Bellville Ted Bonn James R. Burley Ed Fradkin Neil Freeman Herbert R. 1. Grosch Christoph Horstmann Grace Hopper Ernest M. Huber Annual Anniversary on the LesLazar Dinner and Lecture Harvard Mark 1 Theodore C.M. 1.0, M.D. May 5.1983 April 14. 1983 William H. Long Richard Davis Mallery Gregory L. Nelson Walter I. Nissen. Jr. lohn Ousterhout WATCH nus SPACE Ted C. Park lean-Claude Peterschmilt FOR THE ANNOUNCE Robert W. Puffer III MENT OF THE SPRING Dorothy E. Rowe LECTURE SERIES Donald G. Seydel John F. Shoch William D. & Carole K. Strecker IN THE NEXT REPOm: Thomas A. Susie ".....[)r. Ste phen A. Szygenda r ohn Tartar 1irofessor DVR Vithal 0000000000000001 A newsbrief of the collection 1Wo important loans were revealed during Maurice Wilkes' recent gal lery talk on his reminiscences of early computing: a mercury mem ory tank cover and memory driver from Cambridge University's EDSAC. The memory driver (pic tured) was used on the mercury delay line memory. Wilkes was the Director of Computation at Cam bridge, where the EDSAC ran its first progTam in May of 1949. The memory driver and the mercury memory tank cover, on loan from the Science Museum. London, are exhibited on the Pioneer Computer Timeline. • .' _ _ ___a...... __ 1.. ..