The Birth and Glory of Swedish Computers
In the late 1950s, Sweden briefly had the world’s fastest computer. And then in the 1960s, they produced computers good enough to beat even the vaunted IBM. An amazing start. Sweden might have been the computing leader of Europe. But unfortunately, that did not last. In this video we look at the birth and glory of Swedish-made computers with the story of Datasaab. I promise to only make one IKEA joke. ## Beginnings It is the 1930s and the Swedish Government is concerned about rising militarism in Europe.
Prime Minister Per Albin Hansson felt that if war did break out, then Sweden had to be self-sufficient in weapons production. They had excellent shipyards and weapons factories. What they lacked was the ability to make warplanes. So in April 1937, the government founded a company called Svenska Aeroplan AktieBolaget, or Saab. The name literally means Swedish Airplanes Company. Creative. Saab took over the aircraft divisions of the Swedish Railroad Works (ASJA) and began building military aircraft. The first of which was the Svenska B-17, a versatile propeller dive bomber that debuted in 1940. Saab eventually produced some 325 of these planes. When the feared War did
break out however, Sweden declared itself neutral. Their cooperation and vast reserves of iron ores spared the country from a German invasion. After World War II, however, we had the Cold War. The Saab aircraft division continued upgrading their capabilities. In the 1950s, Saab would deliver four jet-powered aircrafts, including the Saab 29.
## Getting Word In 1946, several defense organizations within the Swedish government heard reports about the ENIAC. Which the United States had recently declassified. Immediately seeing the possibilities for cryptography and ballistic trajectory calculation, they sought more information. They sponsored a Swedish electrical engineering professor named Stig Ekelöf to go to the United States - one of the first such to go there since the end of the War. Ekelöf met with professors at Philadelphia and Princeton. Then he also attended the famous Moore School Lectures in the summer of 1946, the first ever on the construction of digital computers. He returned to Sweden in September 1946, giving lectures to defense staff and writing articles about ENIAC's "electron brain".
The Swedish government decided that a "general purpose" math machine would have profound benefits not just for its military, but also for Swedish society as a whole. So in 1947, they founded the Swedish Board for Computing. The board sponsored some number of scientists - four, five or six, the number varies - to go to the United States for a year and join various prestigious computer projects at Harvard and other big universities. Then in 1948, the Swedes tried to acquire a computer from the United States. However, export controls imposed after the Berlin Blockade made that impossible. Thus the Board decided to sponsor the construction of two domestic computers built in parallel.
## BARK The first to be completed was BARK, the Binary Arithmetic Automatic Relay Calculator. Finished in February 1950, it made Sweden likely the third country in Europe to develop its own computer after Britain and Germany. Though who’s counting? BARK was equipped with 5,000 electromechanical relays.
These are electrically operated switches that work sort of like proto-transistors. They physically open or shut to break or connect a current. The whole device cost about $100,000 to build in 1950, or about $1.3 million today. Considering that Sweden’s population was just 6.5 million people at the time, this was a significant expense. BARK's construction was led by the scientist Conny Palm, who envisioned its potential use for controlling telecommunications traffic. Palm unfortunately died young in 1951 from a lung edema, a tragic loss. Like with previous punched card systems, various users like university academics and government were open to use BARK if they had the money. The largest user was the military, who used BARK to run ballistics calculations.
Though they didn't get too much out of it. The only way to program BARK was using its 12,000 hole plugboard - something resembling a telephone switchboard. An input method that took way too long. BARK was later dismantled in 1955 to free up space.
## BESK The second finished device was the Binary Electronic Sequential Calculator, or BESK, completed in 1953. BESK used a stored program computer design, inspired by a famous American computer built for the Institute of Advanced Study by a team led by John Von Neumann. One of BESK's lead designers, Erik Stemme, had been a trainee at the Institute of Advanced Study while this computer was being built. So this similarity was not a coincidence. BESK was powered by 2,400 vacuum tubes, a ferrite core memory, and a cathode ray tube memory system. The latter two were older generation memory systems that existed before semiconductor memories or hard disk drives were invented.
Users input programs into the system using a perforated paper tape. And then it spat output back at you using an electric typewriter. The BESK computer was made available as a time-share, meaning that users would buy shifts to program instructions into the computer and have it run calculations. One early use case was numerical weather analysis of area maps and charts - I covered how computers revolutionized such weather prediction methods in the 1950s in an earlier video. Another use was running mathematical models simulating certain scenarios within Sweden's power system. A simulation that previously took two days on older analog calculators took just 15 minutes on BESK.
BESK greatly influenced Scandanvian computing. It was briefly the world's fastest computer. Certainly the fastest in Europe. And the Danish built their own version of the BESK and they named it DASK - which apparently means "slap" in Danish. That's pretty funny.
I touched upon the BESK computer and its fate in an earlier video about Facit, a venerable mechanical calculator company. The government's unwillingness to provide long-term funding to the Swedish Board of Computing eventually led to its members’ disillusionment - allowing Facit to swoop in and hire Stemme and his whole team in 1957. Facit's new "BESK Boys" as they were called built a copy of the BESK for them - the Facit EDB.
But the company failed to capitalize on these talents, and the company’s antique mechanical calculator business was later steamrolled by the advent of electronic calculators. ## Intensifying Calculations In the late 1940s and early 1950s, Saab worked on a series of leading-edge jet aircraft. To design these high-performance aircraft, Saab's designers needed intense computation. They mapped
the plane's entire body and wing with 3D coordinates, and then manually calculated lift and drag forces at each point to see how well the whole design flew. The Saab 32 Lansen design - done in the late 1940s - had a quarter of a million of these coordinate points. Running the calculations for all of them required a team of 30 "calculator girls" with slide rules and Facit mechanical calculators. Then in 1949, Saab recruited a 34-year old mathematician named Börje Langefors. Upon his arrival, he implemented the widespread use of the Finite Element Method.
This is where we split a design into millions of simpler elements and then run individual simulations on each of those elements. Use of the Finite Element Method grew Saab's computational needs by several orders of magnitude. Hungry for compute, Saab first built an electronic differential analyzer. This is an analog device meaning that it used physical objects rather than digital bits to perform a calculation. Here in this case, wheels and discs.
But Saab's designers wanted something more powerful. So they bought two IBM 604 electronic calculators, which used punched cards to run numbers. This and another subsequent IBM computer still failed to handle the task. So in the early 1950s - just after BARK was finished - Langefors reached out to the Swedish Board for Computing to see if Saab can get their own computer. The aforementioned Conny Palm told him that that would not be necessary because BESK - when completed - can fulfill all of Sweden's compute needs with just a third of its time shift.
This turned out to be a dire underestimate, though not an uncommon one for the age. In 1953, after BESK's completion, Saab used a full shift right off the bat. Between 1955 and 1958, Saab by itself used 44% of BESK's total billed time. They stationed four full-time staff nearby, frequently running it overnight starting after 10 PM.
Over time Saab became frustrated by having to share BESK with others. Insert mildly offensive socialism joke here. So Langefors pushed his company leadership to make a computer of their own for internal use. So Saab licensed the blueprints and the expertise of the Swedish Board's team for what is now 3.2 million Swedish crowns a year to build SARA, the name literally means Saab's Arithmetic Machine.
Completed in 1957, SARA was mostly like BESK except it had improved ferrite core memories and used a magnetic tape called Saraband - one of the first in Europe to be implemented as such. The team's experiences building and using it led directly to their next computer. ## SANK and the D2 In 1959, the Sweden Air Force embarked on a new plane - the Saab 37 Viggen. This design was an ambitious one. Earlier planes needed a pilot and a
navigator - the latter dude for radar recon and other navigation duties. With the 37, they wanted to eject that navigator - figuratively, not literally - and have their tasks be handled by a digital computer, presented to the pilot via a heads-up display. Moreover, the Air Force wanted to replace all the older and unreliable analog systems with new digital ones. The computer would have to handle that too - so a single integrated computer runs the whole plane. Can a digital computer be made small, light, and powerful enough to do all that? Yes, thanks to the recent commercialization of the transistor.
A team at Saab led by Viggo Wentzel - separate from the guys running SARA - was commissioned to build this. It was first called SANK, or "Saabs Navigation Computer" (Saabs Automatiska Navigerings Kalkylator). They thankfully renamed it to D2 and first unveiled it to the public in September 1960. D2 weighed about 150 kilograms and can add two six-digit numbers 100,000 times each second. Better than I can do. We can probably call it Europe's first transistorized computer. Though being a prototype one-of-a-kind device, does it count? The D2 showed the Swedish Air Force that it was possible to have a single-seater fighter jet be run by a single computer.
The Air Force eventually commissioned the CK-37 computer from Saab for their plane, though later planes used an American design licensed and manufactured by Saab. And for Saab, it hinted at something even bigger. By 1960, it was clear that the computer was catching on. In 1955, Sweden had just one computer, BESK. Five years later, there were 25. Two years after that, 35. By 1963, 130. So Saab's D2 and SARA teams joined to found a new division called Datasaab to commercialize the D2 design. As Saab's legendary CEO Tryggve Holm once said, "We woke up one day at Saab and found that we were computer manufacturers".
Within the new division, Wentzel took care of hardware. Langefors, the software. Gunnar Lindstrom, a former nuclear physics professor who earlier made vital contributions to the MRI, became CEO. ## The D21 The D21, released in 1962, was Datasaab’s first commercial product - a medium-sized mainframe for the data processing market in Sweden. As I mentioned in another video, computers for engineering or scientific work do a lot of calculations with a lot of precision, as in many decimal places. So they might need additional circuitry for longer registers.
Data processing computers on the other hand do simpler operations on large amounts of data quickly and accurately. So they might use special circuits to shuffle around large amounts of data. The D21 descended from a military-oriented, high-powered engineering design that seemingly made it unsuitable for data processing. But Langefors boldly argued that the D21's fast and simple circuits and prolific memory - as well as a brand new high-level language called Algol-Genius made it suited for both markets.
The first D21 was sold to Sweden's power utility in late 1960. Other earlier customers included city governments, insurance companies, shipyards, and Volvo. ## The D21 Beats IBM The D21 competed with similar offerings from Facit and IBM. The latter was one of the fiercest competitors in IT.
IBM regularly beat domestic European computing companies with good marketing, deep relationships with IT departments, and strong technologies like disk drives. Buyers often wanted IBM at any cost. For example, while SARA was being built, Saab's own finance director asked if it could process admin data. At the time, this did not seem possible - a good way of sorting records had not yet been published - so the SARA team declined saying the computer would not be fast enough.
The finance director bought a slower IBM computer instead, which seems counterintuitive. To compete, Datasaab marketed deep customization and faultless operation. They equated it to flying, and called it "Saab service". They also included uptime guarantees that IBM was often not willing to match. Moreover, they regularly played the "Buy Swedish" card.
Most famously in 1964, the Swedish government did a King Solomon-style bake-off between the D21 and IBM's 1401 for a computer system to be used by the country's registration and tax collection bureau. The D21 proved to be faster and had higher capacity. But the bureau clearly preferred IBM. They still decided to keep it half-IBM half-Datasaab, but upgraded from IBM 1401s to System 360/30s - the lower end of the recently released System 360 line. After the computers arrived, the government's auditors again found that the D21 computers were not only cheaper than the IBM System 360s, they still also performed better. Thanks in part to their Algol-Genius language letting them use the very efficient Quicksort sorting algorithm whose code had only recently been published. Quicksort in turn relied on Algol's recursion ability. The auditors recommended to standardize on just Datasaab computers, which was done in 1970. It was a rare loss for IBM, and one of the underdog Datasaab's greatest wins.
## The Banking Minicomputers Another one of Datasaab's greatest and longest-lasting achievements was its banking terminal business. In the four Nordic countries - Sweden, Norway, Denmark, and Finland - there are what are called "savings banks". Founded with the goal of encouraging saving amongst the people, they are local institutions owned by foundations. For the American viewers, I would probably equate them to credit unions. The savings banks in all the Nordic countries were experiencing rising administrative costs due to more expensive labor costs.
So in 1968, they started a collaborative project to computerize all their savings banks and connect them to a single banking network. The idea being to make employees more productive and reduce costs. Each bank would have a minicomputer, which then powers many individual "terminals" that each teller can use. You can think of a terminal as like a monitor and keyboard that connects to a computer somewhere. This effort was called "Nordisk Spardata", or the "Nordic Terminal Project". Datasaab collaborated with Facit - which provided peripherals like keyboards and printers - and together they bid for the project in December 1968.
Over ten months, Datasaab executives went back and forth trying to convince both the project joint committee as well as their own top executives back at Saab that this was a worthy project. The final deal was signed September 1969, beating out 14 other vendors. The Terminal Project demanded a minicomputer powerful and flexible enough to meet the needs of four different countries' banks. The bid called for different types of monitors, keyboards, and displays.
To meet these needs, they built the D5/20 minicomputer. Due to its limited ferrite core memories, Datasaab had to invent a new custom language called DIL5 which implemented all the Nordic Terminal Project's software in just 8 kilobytes. They delivered over 2,300 of these minicomputers for the project and eventually sold over 6,000 in total to other mutual savings banks, including a few in the United States. Datasaab was one of Europe's biggest minicomputer makers, with 500 workers in 1976 producing 2,000 minicomputers a year. Though sadly,
the banking terminals were not that profitable for the company to produce. ## Saab-Univac The end of Datasaab and its well-regarded businesses came slowly. In 1974, Datasaab took over Facit's old computer business from Electrolux, who themselves earlier acquired the whole company. Datasaab rebranded one of Facit's minicomputers as the Datasaab D15, eventually selling 320 of them. They also kept selling many of Facit's peripherals, which had been one of that business's few bright spots.
But despite making a growing amount of money on service, Datasaab struggled to make money. The company started cutting off bits of itself, starting with the vaunted mainframe line. Their next big mainframe, the D22, was more powerful than its breakthrough D21 but sold only 70 units. So Datasaab halted production of its next D23 mainframe and then in early 1975, sold the whole D20 line to the American computer company Unisys to form a joint venture called Saab-Univac.
The joint venture moniker was, of course, only a fig leaf. A few years later in late 1980, Saab pulled itself entirely out of that business by selling its stake in Saab-Univac to partner Sperry Corporation. ## Spinoff In 1976, Saab finally had enough of the loss-making Datasaab's business prospects. With the aid of the Swedish Government, they merged Datasaab with the computer terminal division of Standard Radio & Telefon AB and then spun it off as its own company - half owned by Saab and half owned by the Swedish government. The expectation was that the new Datasaab can rationalize and find some sort of synergy with its new partners. But they were still unprofitable and in late 1979,
they lost a second major banking terminal contract to Olivetti, despite being the favored incumbent. Partly because of dissatisfaction with the existing Datasaab product. The terminals were apparently not as good as initially hoped. And updating them would have cost a lot of expensive re-engineering. But more significantly because of customer concerns that Datasaab the company wouldn't be around in five years to service the new terminals. That turned out to be a
self-fulfilling prophecy. Losing a massive $100 million deal sent Datasaab into despair. That and an embarrassing export control scandal in which Datasaab sent advanced American chips to the Soviet Union opened the company up to acquisition in 1981 by Ericsson, the telecommunications company. ## Ericsson and End Ericsson acquired Datasaab because they thought there was an opportunity to enter the "paperless office" space. By integrating Datasaab with some existing divisons to create Ericsson Information Systems, Ericsson felt it can challenge IBM, Xerox and Olivetti for the office automation market. They would expand from selling to telecoms to business firms and government agencies.
One lasting legacy of Datasaab under Ericsson was how it collaborated with the Japanese company Omron to build the ATMs used across all the Nordic area. And that is a nice legacy. But in the end, high R&D costs and low sales in light of the IBM PC boom caused the Information Systems division to turn massive losses. In 1985, they lost 806 million Swedish crowns or $113 million. The year after that, they cut that deficit to 284 million crowns, or $40 million.
In 1987, they sold the part of Ericsson Information Systems with Datasaab within it to Nokia, who then sold it to ICL four years later. ## Conclusion I want to thank the former members of Datasaab who clearly treasured the company's history and accomplishments. For a company I scarcely heard of before doing this video, Datasaab is one of the most thoroughly well-documented with five books in the Swedish language. Check out the Datasaab Friends website. One of the reasons why the Swedish Government and Saab built their own computers in the beginning was because it was otherwise impossible to procure the high powered compute they needed. And for that reason, Datasaab flew ahead of its peers early on.
But as it got easier to procure powerful, general-purpose computers from outside vendors particularly from Japan, customers - even the Swedish ones - were hard-pressed to go with the pricier Datasaab, even if it might be more patriotic to do so. An unfortunate end to a promising start.
2025-02-23 05:49
