Newman's Contribution to the Mark 1 Machines

There are two quotes that I have heard used more than once to back up the argument that Max Newman (and Alan Turing) had a more important role in the Mark 1 story than has been described in these pages. The first is used to assert that the SSEM and Manchester Mark 1 were part of Newman's plan to use his Royal Society grant to build a computer, and even to assert that therefore the Mark 1 was Newman's project not Williams'. The second is used to assert that Williams and Kilburn learnt how to use their storage system in a computer from Newman and Turing, and even to assert that therefore their contribution was crucial to the project.

However I believe that both quotes are less compelling evidence than they might at first sight seem :

1. Status of Newman's Royal Society Grant -- October 1948

This quote is from "A Status Report on the Royal Society Computing Machine Laboratory", prepared by Professor Newman for an internal committee meeting of the University of Manchester, Oct. 15th 1948.

"There are two distinct fields of enquiry connected with large automatic computing machines. These are first, the engineering problem of designing the machine, and especially of finding a satisfactory 'memory' unit, or method of storage of information; secondly, the mathematical and logical problems of finding the best use of such machines and investigating their effect on the development of mathematics itself. The original intention was that the laboratory should be concerned mainly with the problems of the second, or mathematical type, and that a machine should therefore be constructed as soon as possible, modelled on one or other of those which were already under construction in 1946. The most suitable type seemed to be that which was being made at Princeton under John von Neumann's direction. The memory unit [the Selectron], which it was hoped to have ready in six months, was very compact in design, and the idea was to buy from the Radio Corporation of America a sufficient quantity to build into our machine.

It was at this time that Professor Williams was appointed to the Chair of Electrical Engineering. He was already engaged on the design of a memory unit, based on different principles from any previously contemplated. Since his work had only recently begun it was not reasonable to expect his storage to be ready for use in less than two years. Nevertheless, such a fortunate opportunity to combine the two sides of the enquiry was too good to be missed, and it was decided to rely on Professor Williams' storage in spite of the delay that appeared to be involved. What has happened, however, is that while the American unit is still not quite ready, owing to unforeseen difficulties, a small prototype of the machine using Professor Williams' storage came into action about three months ago in the Electrical Engineering Laboratory, and is thus the first of these automatic general-purpose computing machines to have actually worked."

This quote seems to imply that Newman decided to rely on Freddie Williams' storage device as soon as Williams was appointed to Manchester, i.e. a few weeks before he formally started in December 1946. This seems surprising given that Newman could have only just returned from visiting the US with the express purpose of looking for a basis on which to construct his proposed computer. He came back with von Neumann's IAS computer as his model, and the Selectron as his storage device, and with no reason to suppose that the Selectron would be severely delayed. If he did decide to rely on Professor Williams' storage by December 1946, why did he not put any money into the Mark 1 project until October 1948? It is highly likely that Williams, having organised his own funding via TRE, was quite happy to proceed without outside help/interference, but the development up to October 1948 seems to have been on a shoestring budget and there is no evidence that Newman seriously pressed Williams with funds.

It seems far more likely that Newman was not converted until the Autumn of 1947, at the earliest, when the Selectron was now seen to be in difficulty, and Williams and Kilburn were showing that they could store 2048 bits on a standard CRT.

The Royal Society grant was approved in the summer of 1946, and provided for "£3,000 a year for five years for salaries, together with the sum of £20,000 to be spent on construction over the same period".

The first time the Royal Society Grant was used to any substantial degree was in the academic year starting October 1948, when the annual wages component was used to fund Alan Turing's salary and some of the cost of the new Electrical Engineering members on the project. However, given that by October 1948 the government were prepared to fully back the building of a commercial machine to Freddie Williams' specification, it is difficult to believe that, if the grant had not existed, money could not have been found from elsewhere to help fund the Electrical Engineering Department's requirements.

In the end a "home" was found for the grant by applying the capital component and the remaining half of the salary component to providing a new building in 1950 to house the Ferranti Mark 1.

2. Williams remembers a lesson on computers from Newman and Turing

This quote is from "Early computers at Manchester University" by F.C. Williams, presented at the Colloquium on the 25th Anniversary of the Stored Program Computer, held at the Royal Society, London, on the 12th November 1974, and published in The Radio and Electronic Engineer, Vol. 45, No. 7 (page 328), July 1975.

"With this store available, the next step was to build a computer around it. Tom Kilburn and I knew nothing about computers, but a lot about circuits. Professor Newman and Mr. A.M. Turing in the Mathematics Department knew a lot about computers and substantially nothing about electronics. They took us by the hand and explained how numbers could live in houses with addresses and how if they did they could be kept track of during a calculation. In addition, Professor Newman had a grant from the Royal Society. The collaboration was fruitful and by 1949 the state reached is shown in Fig. 3."

This paper is very entertaining, and full of memorable quotes. But it is not written to a sufficient standard of academic rigour to support historical assertions. It contains a number of passages that are seriously misleading or just plain wrong.

Consider first the above quote :

Dr. Alan M. Turing did not join the Mathematics Department until September 1948 (three months after the Baby first ran). Tom Kilburn demonstrated that he knew how the Williams-Kilburn CRT storage system would be used in a computer in his report to the TRE management, A Storage System for use with Binary Digital Computing Machines (Progress report issued 1st December 1947), which explained in detail the research carried out in 1947, and explained how the CRT store could be used inside a stored-program computer. The understanding of computers shown in his introduction 1.4. The function of a storage system, and in the accompanying Figure 1.2, is well beyond the concept of numbers living in houses with addresses. (Interestingly he does show the number store and instruction store as separate "for convenience" while acknowledging that in practice the boundary line between them in the main store must be flexible.) The holding of hands could have occurred before Turing came to Manchester; but Tom Kilburn has no memory of this. Williams almost certainly understood the basic computer model (as far as he cared about it!) around the time he started to investigate the CRT storage possibilities. It can be argued that the basic von Neumann computer model is a fairly simple model (excluding the detailed design to realise it), obvious once stated, and even at that time graspable in a day by an intelligent person, certainly someone with some knowledge of (hand) computation. When I specifically asked Tom Kilburn recently to verifiy this assertion he immediately said that he never read the EDVAC report, and that the concept of storage was obvious from Babbage. He did of course attend the lectures given in early 1947 by Turing at NPL, from which he might have learnt something beyond the fact that he did not like the design, based as it was on a Mercury Acoustic Delay Line store!

The above quote says that Newman and Turing knew "substantially nothing about electronics", a harsh judgement given that Turing had produced a detailed specification of the ACE (including detailed circuit diagrams) by 1946.

That "Professor Newman had a grant" was true, but in effect it was not used until it was not required. Given the government funding pledged to Ferranti Ltd. in October 1948, it would have been surprising if they would not have also helped fund any shortfall in salaries to continue the project.

That "The collaboration was fruitful" is true. (But there is no implication that Newman was the senior partner.)

The final phrase "and the state reached by 1949 is shown in Fig. 3" is the first reference to another aspect of the paper that is very misleading for anyone reading the paper without knowledge of the true history. The Figure has caption

"Fig. 3. The 1949 Manchester University machine. (From Nature, 164, p.684, 22nd October 1949.)"
In fact this refers to the Intermediate Version of the Manchester Mark 1 (available for use from April 1949), the shortcomings of which are given considerable emphasis in the paper (in the interests of telling a good story? -- see two-level store joke below), and no direct mention is made of the completed machine of October 1949 -- see next section (6).

Other Incorrect or Misleading statements in the Williams 1975 paper

Apart from the flaws in the quote given in the previous section, there are a number of other errors or seriously misleading statements :

  1. (Talking about the CRT storage research in 1947:) "The store was quickly developed to the state where one tube could store over one thousand digits (see Fig.2)." and Figure 2 has caption:   "Fig.2. Screen of cathode-ray tube store with 32 words of 29 digits.".  The picture actually shows 29 rows of 32 bits, and has lost the top three rows relative to the original photograph (Photograph 1 in Tom's 1947 report), which shows the display for a 1024 bit store. (Note of course that 29*32 = 928 which is < 1000, and a word on a CRT tube would naturally be held in a row not a column.)

  2. "... and computers with storage capacities of from 12 kilodigits to as much as 340 kilodigits were in existence by 1948". -- If they were, they were not working!

  3. (The end of the famous "dance of death" passage describing the first successful run on June 21st 1948:) "This was in July [sic] 1948 and nothing was ever the same again. We knew that only time and effort were needed to make a machine of meaningful size. We doubled our effort immediately by taking on a second technician" -- It took more than one technician to build the "Baby" (e.g. Tom Kilburn and Geoff Tootill) and more than two people to build the Manchester Mark 1!

  4. (Immediately after a couple of quotes are given from the Letter to Nature, the first published report on the successful running of the Baby:) "This machine had a total storage capacity of 32 words each of 31 digits" -- This repeats the typographical error in the Letter (i.e. 31 not 32).

  5. "This machine solved a variety of problems including checking that 31 is the highest factor of 230 - 1 by the clumsiest routine we could devise. This took 52 minutes during which 3.5 million operations were successfully performed" -- This mixes up two of the three examples given in the Letter to Nature: (1) Long division ("For (230 - 1)/31 this took 1½ seconds, ...") and (3) Factorizing an integer ("Thus the highest proper factor of 218 was found by trying in a single routine every integer from 218 - 1 downward, the necessary divisions being done not by long division, but by the primitive process of repeated subtraction of the divisor. Thus about 130,000 numbers were tested, involving some 3.5 million operations. The correct answer was obtained in a 52-minute run.")

  6. The 1949 machine. The paper makes great play of the shortcomings of the Manchester Mark 1 as it existed in its Intermediary version from April 1949 to October 1949, when it was available for use by University departments and Ferranti Ltd., but without paper tape equipment attached and with magnetic drum transfers only possible by manual intervention with the program halted. But it omits to state that by October 1949 the machine was complete, with paper tape added and both input/output and magnetic drum transfers under program control. A reader who did not know the story would reasonably imagine that there was no further state between the April 1949 version and the 1951 machine, i.e. the Ferranti Mark 1. In fact the 1951 machine is not clearly identified as being the first production model of the Ferranti Mark 1 rather than a further iteration of the Manchester Mark 1.

    The only reference to the Ferranti Mark 1 is immediately after the amusing comments relating to the famous letter from Sir Ben Lockspeiser (the Government Chief Scientist) to E. Grundy of Ferranti Ltd., dated Oct. 26th 1948, authorising Ferranti to "proceed on the lines we discussed, namely, to construct an electronic calculating machine to the instructions of Professor F.C. Williams".   Williams says: "From then on events moved quickly to the Mark I machine inaugurated in 1951 upon which machine a nationwide computer service was based. The charge of £20 an hour seemed exorbitant at the time! Collaborative work with Ferranti continued and Mercury and Atlas followed, but this part of the story belongs to Tom Kilburn ..."

    [It must be remembered that Prof. Williams was manager of the University side of the Mark 1 machines, but he was also lecturing and running a University department. Also he had no commitment to computers; as far as he was concerned the Baby was built to prove that "his" store would be suitable for computers in general. He took an active practical interest in the magnetic drum development, which with its electric motor was much closer to his future interests, but once the CRT store was working in late 1947 it was Tom Kilburn who led the computer design and had the full commitment to computers. Even the development of the CRT store in 1947 was mainly done by Tom Kilburn (with the assistance of Arthur Marsh, replaced by Geoff Tootill, all three on secondment from TRE); he was by now an experienced circuit engineer, and working full time on the store, whereas Williams was getting to grips with running the department. To Williams in retrospect the period from October 1948 to February 1951 may have seemed to pass quickly, but I suspect that to the Manchester Mark 1 team (plus the technicians!), and later also the Ferranti personnel involved, it would have been a long hard but exciting slog. From 1951, there was a formal computer group within the Electrical Engineering department, led by Tom Kilburn, now a Senior Lecturer, and Prof. Williams was able to devote all his time to Electrical Engineering.]

    A passage in the paper starts "Since we are celebrating the 25th anniversary of stored program computers, we must examine whether this 1949 machine satisfied the definition of such a device.". There is then a facetious definition of his own and a serious (but not very rigorous) one given by Newman in 1948. The paper proceeds :

    "Regrettably our 1949 machine did not entirely meet his specification. It required human intervention during the solution of a problem.

    The two-level storage I have referred to was indeed on two levels. The electronic store was in the magnetism room and the magnetic store in the room above. Transfers between the two stores were achieved by setting switches, then running to the bottom of the stairs and shouting 'We are ready to receive track 17 on tube 1'. The process was repeated for tube 2 and the machine set working. When the machine wished to disgorge information, it stopped and the reverse process was initiated.

    In view of the trivial nature of this shortcoming and in view of the fact that it was soon removed, it may perhaps be excused."

    Given this emphasis placed on the inadequacies of the machine, supported by Figure 3 showing the 1949 Manchester University machine as having input and magnetic storage done manually, the phrase "it was soon removed", sandwiched as it is between two sarcastic phrases, seems inadequate as the sole reference in the paper to the fact that there was a fully working Manchester Mark 1 in 1949 (by October), the first machine in the world (by some distance) with a classic two level store (i.e. fast random access to both stores -- RAM in units of a word, secondary store in units of a block of words, or 'page'). In fact the phrase "it was soon removed" probably merely refers to the fact that they introduced an intercom between the two rooms!

Yes I know I shouldn't take this paper so seriously. It is great fun, and must have been very enjoyable to listen to. My point is that other people should not take it seriously either, and not use it to support serious academic points unless quotes have good corroboration elsewhere.

Brian Napper


HOME PAGE
Copyright The University of Manchester 1998, 1999