CSIRAC: The Only First-Generation Computer Still in Existence
For decades forgotten in a warehouse, one of the grandfathers of modern computers — and only the fifth to go live — has been restored.
The world was a different place in 1949: vinyl LP records had just been invented, the now-iconic Volkswagen Kombi van debuted on the market, the world’s first jet-powered airliner took its maiden flight, and the People’s Republic of China was proclaimed by Mao Zedong.
But behind the scenes, it was also the beginning of a powerful technological revolution: the dawn of the computer age. Within months of each other, scientists and engineers had created the first stored-memory electronic computers: large, lumbering machines weighing several tonnes and packed with vacuum tubes and miles of copper wiring. Just four of these strange beasts had come to life in Britain and the United States — the ancestors of the devices that went on to transform the world.
And in November 1949, the fifth such electronic marvel — under construction for two years — finally ran its first program. Its name was CSIRAC, and it was built in faraway Australia.
What’s amazing about CSIRAC is that the researchers who built it had little idea of the work going on elsewhere in the world: scientific journals from Britain and the United States often took more than six months to arrive by ship; there was no direct international telephone service to Australia and telegrams still dominated. Consequently, the local research community was often working in ignorance of the latest developments overseas.
Nevertheless, they set about designing, developing and building a first-generation computer. All of the parts — hardware and software — were built locally. And, by a quirk of fate, this ancient leviathan of the computer age, used well into the 1960s before being consigned to a dusty warehouse, has survived into today, while its siblings elsewhere were stripped apart, often for parts, or dumped.
CSIRAC’s remarkable story began in Sydney in 1947, at the Radiophysics Laboratory of CSIR — the Council for Scientific and Industrial Research, the forerunner of Australia’s national science agency, CSIRO (Commonwealth Scientific & Industrial Research Organisation).
The lab, then based at the University of Sydney, had been looking at new ways to tackle large-scale calculations using electronic components and equipment developed for World War II. But converting radar from its wartime application to everyday civilian use required a level of electronic sophistication and complex calculations that became too laborious to do by hand. Even the mechanical calculators used at the time were too slow.
So Australian researchers — like others elsewhere in the world at the time— reasoned that the leap in technology triggered by the war had made it now possible to build a practical large-scale electronic calculator that could be pre-programmed to handle such hefty computational tasks.
So began the grand endeavour to build a massive electronic calculator. Maston Beard, a research engineer at the laboratory, teamed with Trevor Pearcey, a physicist and mathematician who had worked in Britain for many years on the development of shortwave and microwave radar. In 1946, Pearcey began to design a large electronic computation device with a stored memory, which he called an ‘Automatic Computor’.
Each switching element, each register and each acoustic delay line of memory was painstakingly hand-drawn by pencil on large-format plans that later became the machine’s blueprints — reams and reams of them. Finally, in early 1948, construction began; every thermionic valve, relay and circuit was built locally, with Beard in charge of engineering and Pearcey the design.
The moment of truth came in November 1949, when the first test program was run: a long multiplication routine. And it worked. Clunky is may seem to us — with its 2,000 vacuum tubes and labyrinthine array of wiring — to its creators it was a marvel: it was able to operate more than 1,000 times faster than the best mechanical calculators of the time.
The jubilant team called their metal creation the CSIR Mark 1 (which was later renamed CSIRAC, for ‘CSIR Automatic Computer’). When fired up, it covered 40 square metres of floor space, weighed 2.5 tonnes and consumed 30 kilowatts of power.
“We knew we were in at the beginning of something wonderful,” recalls Peter Thorne, one of the young engineers who worked on CSIRAC before it was retired in the 1960s and who went on to become deputy dean of engineering at the University of Melbourne. “The scale of the machine was impressive. It had noise and a buzz about it — it even had a smell about it. CSIRAC used a lot of power, and that meant that the computer room was always warm, even in winter.”
Before CSIRAC, ‘computer’ was a job description — a person who would wrangle equations on a mechanical calculator. Complex calculations, such as those needed for the emerging field of radio astronomy, would first be broken down into multiple parts and distributed to individual ‘computer assistants’ — row after row of mathematics graduates (mostly women) who would tap out equations for hours, sometimes days, to complete a single task.
“CSIRAC was 1,000 times faster than that, so it was like a ‘supercomputer’ in its day,” quips Thorne.
By today’s standards, CSIRAC was Lilliputian. Its main memory, what we would call today its RAM (random access memory), was just 2k, or just 2,000 bytes. Its long-term data storage — its ‘hard drive’, if you will— held a mere 5,000 bytes. And its top clock speed, or the speed at which it conducted calculations, was 1 kHz, or 1,000 cycles per second.
This compares with the first Apple Watch — launched in 2014 and since discontinued — which had 512 million bytes of RAM equivalent, 8 billion bytes of long-term storage and operated at a clock speed of 520 million cycles per second (or 520 MHz).
At the time, though, instruments like CSIRAC were rare, and it took until the 1960s for computers to begin their long and dizzying climb up the scale of Moore’s Law, which has seen computer processors double in speed every two years or so for decades. Hence, CSIRAC had a long operational life, solving challenges for more than 14 years. In that time, it did the same amount of processing that a smartphone today can do in about a minute.
Nevertheless, CSIRAC was still a wonder of the age in the early 1950s. Twenty years before computer monitors were invented, CSIRAC was using cathode-ray tubes — mini television terminals — to display the internal workings of the machine.
It had no mouse and or even floppy disks: instructions were written on punched paper tape, and then fed into the computer via a small feeder wheel. A photo-electric detector would read each line of 12 holes on the spool of paper tape, row by row. An operator would sit on a ponderous grey metal console covered with toggles, switches and meters.
Once the hour-long testing procedure had been completed, and the paper ‘software’ loaded, with the flick a series of switches, CSIRAC would fire up. Its row after row of grey metal cabinets covered with dials, switches and gauges would come alive. Coloured lights, dotted in rows along its panels, would blink on and off as it processed its task.
Inside the cabinets, a jumble of thick wiring, mercury switches and vacuum tubes would do their job. When it was fully operational, CSIRAC had 2,000 vacuum tubes in its innards — the glass-enclosed valves you see in old radios.
“It is a marvel of engineering,” says Thorne. As an 18-year-old student in the 1950s, he started his career in electronics tending to the machine on weekends. “It was designed and built in Australia; the transformers, the meters on the panels, everything. It’s probably the most completely Australian-made computer that’s ever been.”
It seems hard to believe, but this archaic marvel of 1940s technology nevertheless unleashed scientists to tackle problems that had hitherto been considered too difficult or too laborious.
Unlike today, computer memory was precious, so the early computing pioneers could not afford to write software with millions of lines of code, as software programmers do today. They whittled down their instructions to the barest minimum, often competing with each other to see who could write the shortest instruction set.
Among the problems CSIRAC undertook in its long career were evaluating designs for the construction of buildings, such as the Reserve Bank Building in Sydney and ICI House in Nicholson Street, Melbourne — the latter considered an architectural icon of the 1950s.
Early numerical weather forecasting was attempted — although meteorologists soon discovered that a lot more computing power was needed (and still is!) for such complex interactions. In their quieter moments, engineers also created computer games, calculated their mortgages and even tried their hand at a little computer music.
This was in 1951 and, as it now turns out, an historic moment. CSIRAC’s first programmers were Geoffrey Hill and its designer, Trevor Pearcey. Hill came from a musical family and had ‘perfect pitch’, the ability to identify by ear any note and sing a specified musical note at will. It was his musical interest that to led him to wonder if he could program CSIRAC to play a musical melody.
CSIRAC already had a rudimentary speaker, or ‘hooter’ as it was known, that would warn operators when a task had been completed. To program this very basic instrument to play music required more than a little effort: pulses needed to be sent to the speaker with a regular and predictable period, in order to achieve a steady and continuous tone.
This seemingly simple task was complicated by the fact that CSIRAC had delay-line memory, and each memory tube had a different access timing. Along with a 1,000-hertz main frequency and a memory limitation of just 768 ‘words’, CSIRAC made the task of playing even the simplest tune extremely difficult. Even so, Hill realised that with a little effort, rudimentary music could still be played.
And it was. At the computer’s first public exhibition in August 1951, at the inaugural Conference of Automatic Computing Machines in Sydney, CSIRAC played ‘Colonel Bogey’s March’, a popular marching tune from World War II (you can hear a reconstruction here). Hill went on to program CSIRAC to play other musical melodies, mostly from popular songs of the day, including ‘Bonnie Banks’ and ‘The Girl with Flaxen Hair’.
It was only in the 1990s that historians realised this light-hearted musical foray likely made CSIRAC the first computer to play music, according to Paul Doornbusch, a composer and associate dean at the Australian College of the Arts in Melbourne who specialises in electronic music and has researched the past of both CSIRAC and early computer music.
“The pieces were not as musically inspiring as they might have been if composers had been involved,” says Doornbusch. “The achievement was in conceiving of using a computer to make music, as well as the ingenuity required to produce reliable sounds.
“It’s difficult to appreciate today just how skilful these people were; only two of the best programmers ever managed to program CSIRAC to play music.”
In June 1955 — only six years after it first went live — CSIRAC was dismantled. Whereas researchers overseas built upon their early successes and continued to develop first-generation computers and advance what would become a burgeoning field, Australia — after such a promising start — let its early lead slip away.
Thorne, who got to know many of the early pioneers involved, believes the lack of interest from research leaders and the government of the day stalled CSIRAC’s development. “Trevor Pearcey was concerned all his life that Australia had let a great opportunity slip by,” he recalls. “But there was a feeling in Australia at that time that our main future was as in agriculture and mining. And I think that went through the science and political culture of the time.”
It wasn’t all a loss, he says: “Being launched into the digital age at the very beginning meant that we took on board the idea of computing long before many other nations. We didn’t have to wait for computers to be imported — we had a computer here, and it was used to train people, to develop programming techniques, and so Australia had a really good head start.”
So much so that the demand for more computational power for research was outstripping the ageing stalwart’s capability. By 1955, transistors — first invented in 1947 — were becoming sufficiently reliable and economical to make continued development of CSIRAC’s vacuum tube architecture seem outdated. Hence, engineers at the Radiophysics Laboratory decided to offload CSIRAC and begin to build a more powerful, transistor-based computer known as SILLIAC.
Meanwhile, an American company known for its high-quality electric calculators started selling its first stored-program computer, the IBM 701 Electronic Data Processing Machine. Within a few years, it was shipping fully transistorised, second generation, computers … and the rest is history.
Still the only computer in Australia, CSIRAC was packed up and moved to the University of Melbourne, where it went live on 14 June 1956 and continued to do work for research and civilian projects.
In the eight years it operated in Melbourne, more than 700 computing projects were processed. It tackled everyday problems like insurance risk analysis, calculating government drought-relief programs, simulating the operation of power supply for the southern Australian state of Victoria, estimating the growth rate of pine trees for forestry, producing solar position and radiation tables for all Australian capital cities — and even calculating the housing loan repayment schedules of university staff.
Newspapers of the period lauded it as ‘The Electronic Brain’ and a marvel to behold, although Melbourne’s now defunct The Herald described the computer’s rendition of the university’s anthem as “sounding like a refrigerator defrosting in tune”. But the article went on to admit, “as Professor Cherry said yesterday, ‘This machine plays better music than a Wurlitzer can calculate a mathematical problem…’.”
CSIRAC was finally decommissioned in November 1964, finally succumbing to the power of the transistor. IBM was already making more than US$1 billion a year selling mainframe second-generation computers powered by transistors, and computing power had started doubling every two years.
The behemoth was donated to the Museums Victoria, the organisation that operates three major public museums in the state, and sat in a dusty corner of a warehouse for 34 years. Rediscovered in 1998, it was dusted off and restored, and in 2000 became the centrepiece of the technology gallery at the Melbourne Museum.
It remains the oldest surviving first-generation electronic computer still in existence anywhere in the world, complete and in its original state; sadly, while restored in appearance, it is not functional.
But CSIRAC is nevertheless a marvellous reminder of an era long past; an era when the idea that one day anyone could have their own computer at home — much less on their wrist — would have seemed utterly preposterous.
THE FIRST COMPUTERS
The difficulty with establishing a definitive history of computers is agreeing on just what a computer is. Many groups claim the title, and some early electro-mechanical calculators were able to do large-scale calculations as far back as 1941.
It’s also difficult to assign a ranking to the operational dates of many of these first-generation computers because much depends on the definition of ‘operational’. Does it mean when the first test program was run, or when the computer started routine operation? Early electronic computers were room-sized custom-built pioneering research projects. Once the basic operations were shown to be sound, the machines were continuously improved.
However, applying today’s definition of a digital computer — an all-electronic machine capable of calculating operations, where the data and instructions are held in rewritable memory — then the chronology of the first computers to go ‘live’ is this:
April 1949: The Mark I, built at Manchester University in Britain, runs its first stored program.
May 1949: EDSAC, built at Cambridge University in Britain, goes live and runs its first stored program.
August 1949: BINAC, built by the Electronic Control Company of the United States (consisting of ex-University of Pennsylvania engineers), runs its first stored program.
September 1949: The Harvard Mark III, designed and built at Harvard University in the United States, runs its first stored program.
November 1949: CSIRAC goes live, running its first program — a long multiplication routine.