|Columbia University Computing History|
Lentz said of the 610, "A novel approach to computer programming and control, used in the IBM 610 computer, allows the solution of complex problems by an operator whose only previous experience with computing has been the desk calculator. The machine's command structure is designed so that the operator can at all times communicate with the computer by a series of short sentence-type instructions closely resembling the steps of manual arithmetic solution. A type of floating-decimal operation called the 'auto-point' mode permits entry of data into storage locations with automatic positioning of the decimal point, without elaborate programming. The decimal point is automatically re-positioned during subsequent computation" (Reference 1).
Users said (Reference 2) the machine was affordable, reliable (95% uptime was typical), easy to program (it was one of the first — if not the first — computer that was programmable symbolically from a keyboard), handled floating-point arithmetic naturally, and did not require air conditioning or special power. Some, however, faulted it for its execution speed (e.g. 20 seconds to calculate a sine). But as Brennan says, "Far ahead of its time conceptually, the 610 presaged 'on-line' direct communication between individual and computer." When the 610 was discontinued (it was technologically obsolete from the start, due to the long delay in getting it to market), most sites replaced it with a 1620.
IBM produced several other personal computers in later years, including the 5100 and the CS-9000 before
finally releasing its world-conquering PC in
1981 (the CS-9000 was ready before the PC but announced after it).
For reasons lost in time, the construction of the first prototypes was farmed out to Burroughs/ElectroData in Pasadena, CA, which also made contributions to the design. In May 2004, I received the following comments from John C. Alrich, who was on the 610 design team at Burroughs, and worked with Lentz for 12-18 months in Pasadena on the project:
I was on the design team from Burroughs. In fact, I had a patent on part of the drum design. John was indeed the architect but Burroughs, Pasadena, played no small part in the design and building of several of the prototypes. The only printed data I have from that project is my patent which was filled 4/14/55 and issued 9/17/57, so April '55 must have been in the middle of our design phase in Pasadena. I have no other documents. I do remember Herb Grosch coming out and looking at the machine when it was well along [John, with Jack Palmer, also of IBM, had been up half the night getting the square root function to work for this demo; the 610 was the first IBM product with built-in square root capability*.
I don't remember whether we were part of Burroughs yet or still a subsidiary of Consolidated Electrodynamics Corporation, called ElectroData. CEC manufactured mass spectrometers and our first computer was designed to invert large matrices that were used in compound analysis. The guy that pushed CEC into the computer business was Clifford Berry, who designed mass spectrometers, and who — are you ready for this — got his Ph.D. under Atanasoff before World War II and worked with Atanasoff on his first computer there at the University! Cliff did not work on our first computer, called the Datatron 201, but continued with designing mass spectromters. I think Cliff died in the late fifties at an early age.
John's design was radically different from the CEC/von Neumann design I was familiar with in that the circuitry was dynamic rather than static; i.e. he used free-running multivibrators rather than static flip-flops for his logic. He did not think ff's were stable! I can go into more of that later.
The other strange thing in the design (to me at least) was that the 610 was essentially a Turing machine; i.e., in principle, it had infinite capacity for input data and for intermediate and final output data. The means, of course, was punched paper tape, both working, as I recall, at 18 characters/second! The small plated drum was also used to store intermediate results. John used lots of wire relays in his design too.
Why I was assigned to the project by L.P. Robinson (Robbie), I'll never know. I was not a circuit person although, from 1951-52 I worked under a brilliant mathematician, Ernst Selmer, who was the #2 mathematician in Norway and was working with von Neumann's group before he came west to teach at Cal Tech for a year or so. So I knew logic design pretty well (I designed the Floating Point Control for the Datatron in 1957, the most satisfactory piece of design I did in a career of 40 years).
It was interesting to read that IBM made 180 units, about which I have only two comments:
- Because of the dynamic circuitry, if the clock lost synchronization, you couldn't hold a fixed image on a 'scope screen to do any debugging; and
- When this happened, Lentz was one of the few people in the world who could analyze the trouble and fix it.
I wonder how IBM's Field Service managed? On re-thinking about the 610, I still find it an enigma. There were a lot of clever ideas in it, mostly John's, but I think John went up the wrong limb of the computer evolution tree. In principle, his machine could solve any mathematics problem that could be solved in finite time but by using relay trees and paper-tape I/O, the speed of execution was intolerably slow, even by 1955 standards.
|*||The Aberdeen Relay Calculator (1944) did square roots too, but it was not a product on the open market.|
John Alrich comments (June 2004): "In many ways the 610 was unique or nearly unique for its day or any other day. One attribute in particular was the method of numerical encoding. As I recall, each word was fifteen digits long using pulse position encoding. That is, each of the fifteen digits was twelve serial slots long. Depending upon where a pulse or pulses appeared within each digit determined the value of that digit, the sign of the word and the decimal location. Therefore the serial display was rather simple — a CRT with a single modulated beam. A transparent engraved reticle, with 180 small slots, placed in front of the CRT allowed the user to read out immediately the numeric value of the word being displayed.
"I think with pulse-position modulation it becomes fairly obvious how certain arithmetic operations are performed. Since the 610 was non-Boolean, arithmetic operations were generally done by counting rather than with Boolean logic. Using the drum for temporary storage, during an add, for example, the addend and the augend are simultaneously available in serial form. The add operation probably was done something like the following: The addend and the augend were read off the drum in synchronization serially, least significant digit first. In between clocks on the drum (enclosing a full word), each digit of the augend was counted down at a high rate to zero while each digit of the addend was counted up, with carry noted for inclusion while counting the next digit. In this way the sum was determined within one word time of the drum revolution. Subtraction could be performed in a similiar way replacing the carry with a borrow; multiplication, division and square-root were, of course, more complex."
John relates that the people at Burroughs called the 610 the CADET
("Can't Add, Doesn't Even Try"), the same term used by IBMers for the
1620. For more about John's experiences at
Offsite Links (valid as of 29 April 2009):