Sputnik, spaghetti and the IBM SPACE machine
The 50th anniversary of the 1401
The fall of 1957 was a low point not only for the United States, but for another high-minded world power: IBM.
As the US looked up at Sputnik, without a satellite of its own, IBM was haunted by a rival machine known as the Gamma 3. Built by an upstart French outfit, the Gamma had trumped Blue Blue's fledgling computer tech - not to mention its European sales - and the company lacked even the blueprint for a response. In the fall of 1957, the International Business Machines Corporation had no major business machine in development.
But just as the US would reclaim lost ground with the Mercury space program, IBM would battle back with the 1401, a seminal stored-program machine fated to become the most popular computer of the 1960s.
This year marks the machine's 50th anniversary, and last week, at Silicon Valley's Computer History Museum, three of the system's founding fathers gathered for a night of nostalgia. Introducing the venerable trio - chief architect Francis Underwood, project head Charles Branscomb and planning manager Sheldon Jacobs - current Big Blue marketing boss John Iwata called them "IBM's version of the Mercury astronauts".
WWAM. Bam. No thank you, Ma'am
Shipped in 1953 by a French outfit known as Compagnie des Machines Bull, the Gamma 3 was an early electronic computer that served as both accounting machine and general purpose calculator. Big Blue had dominated the accounting machine market for decades, but according to Charles Branscomb - an exec in IBM's Endicott, New York lab at the time - its technology had suddenly fallen behind.
"It was a wake-up call. The Gamma 3 was a calculator that actually attached to an accounting machine, and the IBM Europe people were very concerned," Branscomb says. "They came to the corporation and alerted the corporation that there was a competitive threat - and that something had to be done about it."
Big Blue's initial response was to fashion a transistor-based system known as the Worldwide Accounting Machine, WWAM for short. Its processor was designed at IBM's French lab in Paris, its punchcard reader and printer at a German lab outside Stuttgart.
In 1957, Branscomb was appointed area manager for accounting machines at the Endicott lab, and his first task was to review the WWAM's progress. That October, he flew to France for a meeting with the product manager overseeing the processor design. And as Branscomb sat down, the Frenchman couldn't help but mention the satellite the Soviets had launched the day before.
"The Russians put up Sputnik yesterday, and it goes around the world saying 'Beep, beep, beep,'" the Frenchman said. "Then, when it gets to the US, it says 'Ha, ha, ha.'"
The IBM 1401 (right), with card reader and printer
The irony is that the IBM - including the Frenchman himself - was in much the same position as the Sputniked Americans. After Paris, Branscomb flew to Germany, and he soon told the corporation that the WWAM project should be terminated. Like so many accounting machines that came before it, the WWAM was programmed through plugboard.
"[The European labs] had done some creative work. They had tried to make the best use of a control panel, but it was still a control panel machine and technology was on the move and my conclusion was that it was not going to get the job done," Branscomb remembers.
"Here it was, late third, early fourth quarter 1957, and IBM's largest revenue base had no machine in development."
Except that it did. To replace the WWAM, IBM turned to a system design sketched out by a single man working in Branscomb's Endicott lab. "I was called the chief architect of the 1401," Francis Underwood says. "But as a matter of fact, that was a made-up title. It wasn't official. It implies there were lots of other non-chief architects floating around. But they're weren't any."
The line is played for laughs. And Underwood goes on to praise the contributions of countless other IBMers in actually getting the system built. But in pegging himself the sole architect, he isn't far from the truth. Branscomb calls him the project's "creative fire".
After a stint in the Navy and a 75-cents-an-hour job as a tool maker in Washington, DC, Underwood was hired by IBM as a "customer engineer," someone who would answer the needs of the contracted clientele. But in his spare time, he designed machines of his own, "breaking out the old drawing board" when he came home in the evenings.
Eventually, he shared two designs to his supervisor: one for calculating keypunch, a device that could not only punch holes in the machine-readable data cards of the day but also make calculations on its own, and another for a newfangled interpreter, something that could both read cards and print to them. The supervisor sat on both. But within a week, Underwood had arranged his own interview with the lab in Endicott, and that nameless supervisor was soon left behind.
In Endicott, Underwood would join IBM's advanced systems development department, a group of engineers and mathematicians charged with predicting the future. "We were supposed to figure out what was coming ten years out," Underwood says. "But what we really did was screw around - and pretend we knew what we were talking about."
If Underwood screwed around, that wasn't all he did. Under the advanced systems aegis, he would lay out the design for what would become the 1401. Appropriately enough, he called it the SPACE project, short for Stored-Program Accounting and Calculating Equipment. Like the Gamma 3, it combined accounting machine with calculator. But unlike the Gamma - and unlike the WWAM - it didn't use a plugboard.
"We weren't interested in building an accounting machine," Branscomb says. "We wanted to take people to a whole new world."
No spaghetti for you
For decades, electromechanical accounting machines had processed data stored on punchcards - literally pieces of paper with holes in them. And since the early years of the 20th century, such machines had been programmed via plugboards, arrays of wires and wire-jacks.
"We have an image here of a fairly simple control panel," Branscomb deadpans, pointing to a mass of interwoven wires on a screen behind him. "They looked like a bunch of spaghetti, but this is how you controlled the machine in those days."
When Underwood laid eyes on the WWAM's plugboard - which accounted for nearly half its cost - he was, in his own words, appalled. "It was huge - it had to be," he says. "And all the electronics that drove the panel did nothing but that. They didn't help the customer solve his problem. They just drove the control panel."
The SPACE Machine would take a very different path. "I said 'From now on, no more control panels. We can use the money in a much better way,'" Underwood explains. "We could go to the kind of technology we saw in the large scale binary computers, go to a stored program." In this way, programming could be treated like native logic. "It could be manipulated and executed at high-speed."
Underwood calls the SPACE machine "my own little project. It could sit there and I could hug and pat it, slice it and dice it, and do anything I wanted to do. Nobody was telling me what to do." But in the wake of the WWAM's demise, it would become much more.
Joined by the 1402 (a card reader) and the 1403 (a printer), the 1401 was announced to the world on October 5, 1959, just two years - to the day - from Branscomb's Paris Sputnik meeting. The first unit shipped in September of the following year.
Clock speed: 87 kilohertz
Weighing four tons and consuming 13 kilowatts of power, 1401 machine included 500,000 separate components. Borrowing the WWAM's data path design, it was one of IBM's earliest transistor-based machines, with discrete germanium alloy-junction transistors designed by Big Blue and eventually built by Texas Instruments.
Made up of 2,300 printed circuit cards - each about 2.6-inches wide and 4.5-inches long - the core processor ran at 87 kilohertz. Core memory was no more than a stack of donut-shaped magnets circling tiny wire threads, and it stored somewhere between 1,400 to 16,000 positions, each holding 8 bits of data.
The 1403, the 1402, and the 1401 - with a tape drive
Machine-level programming was handled through a language dubbed Autocoder, but compilers were soon added for Cobol and Fortran. Though there was no operating system per se, Big Blue did provide a library of input and output utilities, and for later units equipped with tape drives, the company offered sort and merge programs as well.
The typical 1401 sold for around $500,000 - roughly $3.4m in today's dollars. But you could rent one for a mere $6,500 a month.
For some, Fran Underwood's SPACE machine wasn’t the most important IBM system of the 60s. The System/360 mainframe played Apollo to its Mercury. But as Charles Branscomb puts it: "The 1401 made a huge contribution to the 360. It paid for it."
In 2004, IBM research and design veteran Robert Garner set out to restore a 1401 to working order. And five years later, he and his team have not one but two up and running at the History Museum.
The 1402 card reader
The machines have a cycle time of roughly 11.5 microseconds, and they need about 50 cycles to add a 20 digit number. That makes them about a million times slower than a modern PC. But in some respects, Fran Underwood says, the 1401 is still the superior machine.
With his SPACE machine, Underwood remembers, you could calculate the powers of 2 with a mere nine instructions. "Today, you'd need tens of thousands," he says. "That's terrible." ®