Birth of the Desktop ComputerThis… Is just a little piece of fiction that I’ve written some time ago, as a piece of Alternate History Worldbuilding and Thought Experiment. I don’t think there is anyone who has written something like this before…

The Nixdorf 8810 Schreibtischcomputer

Birth of the Desktop Computer

It was the year 1976, when Heinz Nixdorf, founder of Nixdorf Computer GmbH in Paderborn, returned to the United States on a journey to visit some old friends from his days at Remington Rand Corp and do business.

At the time, Nixdorf Computer was one of the many competitors of IBM, and their main competitor in Germany, when it came to mid range computer systems. Its line of Nixdorf 88xx data systems, from microcomputers over bank terminals to conventional terminals, was successful in the market and would, by 1979, push the company over the billion DM in sales.

Hoping to further expand his company’s sales and product line, he first arrived in Cambridge, where he visited MIT, hopong to recruit new talent for his company. Here he met with people of the Artificial Intelligence Laboratory, building excellent rapport with a few programmers and hackers there. Compared to the owners of companies in the US, this might be surprising, but one had to remember that Nixdorf was a computer engineer himself and had built the first product of his company himself.

But it was not only the Artificial Intelligence Lab, where he stuck up friendships, but also the Dynamic Modelling Group, where he met Professor J.C.R. Licklider, who was important to the history of computing in the US by his varied and sometimes instrumental ideas.

After his visit at MIT, Nixdorf continued to visit Xerox in Rochester, New York, intending to look at some new photocopier systems offered by the company for his own company. In Rochester, he had the first glimpse at the Xerox Alto, which had only recently arrived from Palo Alto, and had not gathered much interest in the managers of Xerox.

The Alto managed to get Nixdorf’s attention, and he got a brief presentation of the computer and almost immediately saw the potential. Previously, Nixdorf had not been concerned with the new microcomputers developed in the early to mid 70s, such as the Altair 8800, which he believed were mere toys. Now, however, he began to change his mind.

While negotiating for a number of copiers, he managed to visit PARC as part of the deal, to look at the machine that had caught his interest.

And so Nixdorf arrived at Xerox PARC in June 1976 and was invited into the facility. His knowledge as a computer engineer made him ask the right questions on the Alto to get a more in-depth look at it, including being allowed to try it out for himself. He asked enough questions to get well-informed about the computer, and the belief that small computers died with the answers he received.

During the visit at PARC and the return to Germany, Nixdorf became convinced the Alto would be a gold mine for Xerox and that the company would release the computer into the market within the next one or two years, possibly pushing Nixdorf and his 88xx system out of the market. He knew the Alto would need to be produced cheaper than it was currently to sell on the open market. Xerox was surely already working on that problem.

So, Nixdorf needed to be on top of the problem and offer a comparable, maybe even superior product to the Alto around the same time, or even just a short bit after the Alto became available.

To this end, Nixdorf called in his best engineers and presented them with the problem of a 16 bit computer with a fully graphical user interface, to be controlled by a pointing device, called a mouse, in addition to the more usual keyboard. When the engineers said something like that would be impossible to build, he showed them photographs of the machine. He could convince his engineers that Xerox would corner the market for the decentralised data systems Nixdorf was creating, with the Alto, in addition to offering many more potential applications.

Off the top of his head, Nixdorf did envision a spreadsheet program, presentation software, graphical database systems and a few others.

As such, the 8810 project began.

The Nixdorf 8810 Scheibtischcomputer

Hardware

The first problem was the CPU. The Xerox Alto was a 16 bit microcode system, implemented in TTL logic, something Nixdorf believed to be a thing of the past. With the new highly integrated CPUs on the market already, the Intel 8080 or the Motorola 6800, discreetly built CPUs were simply not viable anymore. 8 bit machines were also likely to be obsolete in the near future, as more powerful 16 bit and even 32 bit CPUs were on the horizon. For 16 bit CPUs, Intel had just announced the 8086.

However, Nixdorf was not sold on Intel and its 8086. Yes, it would be a processor available in the next year, but Nixdorf believed he should look around for an alternative. Some rumours in the industry were noting Motorola had been quiet on the 16 bit market, when other companies had already introduced their own, so Nixdorf got into contact with the company.

Not believing Motorola or any other company could deliver a 16 bit processor that could support what he was looking for, Nixdorf used his connections to the University of Paderborn to look into any other possibilities. Nixdorf was in luck, as he ran across Michael Dolkind, a former IBM engineer who had only recently begun working at the University, after returning to his hometown to help take care of his ailing parents.

Dolkind had worked on the IBM 801 project, which had run from 1970 to 1974, and had aimed to develop a processor that could run code quickly to route telephone calls. To this end, the IBM 801 team had come to the realisation that they could drop everything not necessary for doing its job. This had resulted in a very fast processor with a reduced set of instructions that could be run within a single clock cycle. Dolkind had also heard about the CDC 6600, a mainframe computer that used a similar technique to speed up its processing.

Nixdorf was interested in the idea and invited Dolkind to his company to give a presentation about the IBM 801 to his engineers. Nixdorfs own engineers were fascinated by the idea and pulled Dolkind into a new project to develop a new processor.

The result can be called the first 32 bit RISC processor, even though the term RISC, Reduced Instruction Set Computing, had yet to be created. Internally named the 3200, the processor was a 32 bit load-store optimised RISC processor with a register to register architecture, using 32 32 bit registers that forwent the use of an accumulator. It was equipped with a 3 stage pipeline that allowed it to execute a single instruction per clock cycle. To connect to the outside world, it had a 32 bit address and data bus.

Its instruction set was reduced to an absolute minimum, but allowed various addressing modes to deal with data, such as direct, indirect and relative addressing.

The project was completed by late 1976, and a prototype, built in TTL logic, worked as advertised, operating at 2 MHz. All that was needed now, was a partner to create a version in silicon.

Nixdorf looked around, asking friends at Siemens, Bosch and Phillips if they were interested in a cooperation to produce the 3200. In the end, it was Phillips who was interested in the project, especially after Nixdorf noted he was also looking for a high resolution graphical colour display for his new computer. And Phillips was in the business of building TVs.

This resulted in the design and production of the Phillips MC-3200 by mid 1978, produced in the Phillips semiconductor plant in Hamburg. It was the first major processor to forgo the DIL chip design and use a 72 pin PLCC.

The pin compatible Phillips MC-3210 followed in 1980, integrating 32 and 64 bit floating point units with 4 64 bit registers that could also act like 8 32 bit registers. The MC-3220 replaced it by 1986, which supported IEEE 754 floating point numbers.

Seeing something of a market, Phillips also began to produce dynamic RAM chips themselves, compatible with the 4116, 4132, 4432, 4164 and 4464 designs.

With the MC-3200 as the main processor for the 8810, much of the hardware design followed quickly. The 8810 would use at least 256kB of dynamic RAM in initially four dedicated memory modules, which the user could exchange later for larger modules. These modules would integrate the regeneration circuit needed to keep the memory contents viable.

The idea was also that the computer’s main operating system would reside in ROM, enabling quick boot times and preventing problems later on, like damaged magnetic memory. Some main programs would also initially reside in ROM, utilising specialised ROM modules for the software.

For magnetic storage, the 8810 initially was meant to carry a single 8 inch floppy disk, which was later joined by one or two 5.25 inch floppies.

The 8810 also took a page from the Altair 8800 and came with several extension slots to take additional modules to extend the capabilities of the 8810. These were two 32 bit slots, three 16 bit slots and four 8 bit slots, though the higher bit slots were designed to be pin compatible with the lower bit slots, so the 8810 could take nine 8 bit modules or five 16 bit modules.

For data entry, the 8810 used a conventional keyboard, though it was separate from the display, as well as a simple two button mouse, using a rubberised steel ball and simple incremental encoders to keep track of the position. Additionally, the 8810 could work with a light pen and other interfaces.

Graphics

Since the Nixdorf 8810 was meant to compete with the Xerox Alto, there was a clear need for advanced graphics capabilities. This in turn meant it could not use a conventional terminal, even if it would later be an option. Existing graphics terminals were out of the question, as they would be too expensive for the aimed low cost of the 8810.

Instead, Nixdorf’s engineers needed to find a way to create a way to make the 8810 itself display graphics on the display.

Early on, they came down on a display that had more in common with a conventional TV screen in landscape orientation, rather than the Alto’s portrait orientation. That way people would not be put off by the screen of the computer itself. The screen would be 14 inches diagonally, with 768 by 512 pixels, rather than the Alto’s 606 by 808 pixels. They correctly assumed it would be easier to scroll the screen up and down, rather than try to make the screen appear like a sheet of paper as with the Alto.

This alone meant the screen would need at least 48kB of screen memory for a two colour bitmapped screen, as with the Alto, and put the burden of filling this screen memory to the processor.

Once it became clear that the 8810 would use the MC-3200 as processor, the idea came up to use a second MC-3200 for the display. Since the MC-3200 was fast and had a very simple instruction set, its program could be changed on the fly, say to operate in bitmapped graphics mode, or to display more than two colours.

To this end, the 8810 team developed the first dedicated graphics card, the Nixdorf 8815, which would fit into one of the 8810’s 32 bit extension slots. The 8815 would be equipped with a MC-3200, 64kB of graphics memory, 16kB of ROM for initially two graphics modes, and 16kB of dynamic RAM to dynamically take other graphics modes. There were also hardware character generators in ROM, as well as some memory for software based character generation.

The two integrated graphics modes used the first 48kB of the graphics memory to store bit mapped graphics of the main screen, and would be used to generate the image for the display. 

The remaining 16kB of the graphics memory would be used for what the Nixdorf engineers called Farbräume, Colour Spaces. Each Colour Space was 8 byte in size and contained x and y coordinates of 12 bits each, the x and y size of the Colour Space in 12 bits each, as well as 8 bits of background colour and 8 bits of foreground colour. Each of the Colour Spaces could overlap one another, with the layering of the colours determined by the location of the Colour Space in memory. The Colour Spaces could be as large as 768 by 512 pixels or as small as 1 by 2 pixels.

In this Colour Space graphics mode, the display could show up to 2048 Colour Spaces with either 128 colours and 128 grey scales as fore and background or up to 256 grey scales. Later on cheaper memory allowed more and more Colour Spaces to be supported this way. In the 1990s, the size of graphics memory allowed colour bitmapped graphics.

The first sample design of the 8815 was finished by late 1977, using another TTL based 3200 prototype.

Networking

While the 8810 was initially imagined to be a stand alone system for single users, it quickly became obvious that it would be hard to exchange documents and the like through exchanging floppy disks.

Nixdorf and his engineers were already aware of ARPANET, and the idea was to build something like ARPANET for the 8810 and any family of systems developed on its design. Looking for documentation, they quickly stumbled over a paper by Robert Metcalfe and David Boggs, “Ethernet: Distributed Packet Switching for Local Computer Networks”. To make things interesting, Metcalfe worked for Xerox and developed Ethernet for the company.

In a move that surprised many, Nixdorf approached Xerox to licence the patent for Ethernet, so Nixdorf could connect their machines into a decentralised network and use the new Xerox 9700 laser printer. Xerox, not knowing that Nixdorf was developing a competitor to the Alto, did give Nixdorf a single payment licence for Ethernet, allowing Nixdorf to develop and build their own Ethernet networking hardware for the 8810.

Software

With the main processor sorted and the graphics work underway, the 8810 project team began working on the software needed for the project by early 1977. Initially, this meant the most basic operating system, which could later be extended by the graphical user interface and other programs.

The project to create the main kernel for the 8810 fell on a pair of University of Paderborn students working at Nixdorf, Alexander Wilhelms and Thorsten Schmidt. At the time, both students were interested in UNIX of Bell Labs, but could not get any source code for that operating system. Instead, all they could get were manuals for UNIX. They were not deterred from trying to create their own version of UNIX.

Originally, they wanted to write their UNIX in C, but were soon disappointed that the language was missing a few things they had seen in documentation about Small Talk, which was used on the Xerox Alto. As such, their first aim was to extend C with object oriented programming, creating the programming language D, which they then used to program the 1978 UNIX compatible Pader Kernel. Initially, the Pader Kernel was running on the Nixdorf 820 minicomputer, before it was ported to the 3200 prototype processor, and got it working by mid 1977.

From there, they began to extend the main kernel into a full basic operating system by reverse engineering other support programs used by UNIX, including a full compiler for D, allowing the Pader Kernel to self host on the 3200 and following 8810 systems.

They soon programmed full interpreter environments for BASIC, LISP and FORTH, before developing a simple object oriented version of BASIC, ObjektBASIC, all of which would be pre installed as part of the 8810 software package. Compilers for FORTRAN and COBOL would follow later on.

With the 8815 ready for use, the Graphical User Interface quickly followed, as Wilhelms and Schmidt had prepared for it. By mid 1978, they had the bare bones ready, with icons, a mouse pointer, overlapping windows, dynamic scrolling and the ability to change the size of windows. Just in time for the first prototype 8810 with the first batch of prototype Phillips MC-3200.

The GUI was a WIMP (Windows, Icons, Menus, Pointer) design, somewhat reminiscent of the Xerox Alto, but showed many differences in general design. The main initial programs were a text editor, a terminal emulator, and a simple graphics program, followed by a graphical file manager, and various modules and dialogues that could be used by other programs.

With Wilhelms and Schmidt working on the main Operating System and improving it, Nixdorf hired more programmers and gave them their own projects.

Among them were programs to write documents, edit images, and, most importantly, to use a computerised version of spreadsheets for data entry and direct manipulation of data in one go. In essence, Nixdorf TabellenKalkül was a direct contemporary to VisiCalc on the Apple II, designed independently.

By the end of 1978, the Nixdorf Grafisches Betriebssystem, NGB, had passed its first major milestone and began to enter operational use at Nixdorf GmbH, connected by Ethernet networking.

Networking itself quickly developed its own new programs, such as EMail and file sharing between computers. Additionally, the design of the 8810 and the Pader Kernel itself allowed the quick development of the Nixdorf 8830 Netzwerkspeicher, a 8810 with hard drives attached for central file storage.

Several months and intense use of the 8810 allowed the system to be tested in an operational environment and implement updates and suggestions to make NGB and its programs easier to use.

Internally, the entire initial NGB OS was stored in ROM, but newer versions of the software, from the kernel to its components, could be loaded from floppy and later on hard drive. The user would replace the ROMs, and with a EPROM writer module, they would even be written to EPROM to replace existing ROMs.

Extension cards themselves were expected to carry their own ROMs, containing both drivers and programs to use the extension card.

Presentation

For the first presentation of the Nixdorf 8810 Scheibtischcomputer and the 8830 Netzwerkspeicher, Nixdorf decided to go with a bang and present it to the public at the Winter CES 1979 in January in Las Vegas.

It was the year when Texas Instruments presented their own TI-99/4 and Bill Gates presented a version of BASIC for the Apple II. The Nixdorf 8810 however overshadowed these two introductions, and it certainly hit the representatives of Xerox and, perhaps more important, IBM in the face.

Here was a computer that was quick and easy to use, networked with other computers of the same time, allowing to write emails, share files, write documents or work with spreadsheets.

While the representatives of Nixdorf were quick to note that all these computers were still prototypes, the finished product would cost just 2500 US$.

To say the 8810 had a big impact on the desktop computer development of the early 1980s would be an understatement. Due to the 8810, Xerox’s management took a very hard look at the Alto, which they had ignored until the Winter CES 1979, and pressured PARC and Xerox’s Systems Development Department to develop similar capabilities, resulting in the Xerox Star from 1980.

At IBM, President John Opel assigned William C. Lowe to the new Entry Level Systems unit in Boca Raton, Florida, to develop something to counter the 8810, resulting in the IBM 5510 Personal Computer in 1980.

At the convention, Nixdorf himself met with two men he hoped would help him offer the 8810 and 8830 in the US. One was Jack Tramiel of Commodore, but Nixdorf had an intense dislike of the man from the get go, largely due to his abrasive personality, something returned by Tramiel.

The other man was Ken Olsen of the Digital Computer Corporation, DEC. Other than Tramiel, Olsen has an easy time connecting to Nixdorf, since they both started out similarly with their own companies. In the end, DEC would get the first licence for producing the 8810 and began developing software to connect almost seamlessly with DEC’s lineup of PDP mini and mainframe computers.

Later on Nixdorf would extend the licence deal to more companies within Europe, to buffer against the large demand for the 8810.

With the Winter CES 1979 under their belt and the DEC deal, Nixdorf waited until May 1979 to begin offering the 8810 for purchase, initially to German customers. This allowed them to do more bug fixing of the software and build up a stock of the hardware. But even then, the orders overwhelmed Nixdorf for the 8810 and the 8830, as several large German companies ordered the systems to equip entire offices.

In the US the 8810 and 8830 likewise flew off the shelves of DEC, mostly to smaller and medium companies, as well as universities, who were looking favourably to the comparably cheap and capable computer.

One 8810 found its way into the hands of Richard Stalman, who was enamoured by the capable machine and flew to Germany and went to talk directly with Nixdorf and the teams around the 8810. By 1983, the Pader Kernel would be part of the GNU Project, and much of the software developed for the 8810 would become open source.

Legacy

The legacy of the 8810 is undisputed.

It gave the world the first 32 bit RISC processor and the first viable graphics workstation and capable desktop computer a person could own by themselves.

Sure, it was more expensive than the 8 and 16 bit microcomputers that would rule the low cost market in the 1980s, but it forced large companies, like IBM and Xerox, to follow a trend set by a small German company.

It is unknown if the modern graphical user interface would have been possible without the example of the 8810, or how the operating system market would have looked without the free and open source software that can trace its success back to the 8810 and the Nixdorf Grafisches Betriebssystem.

The IBM 5510 was the only capable competitor to the 8810, even if it initially lacked many of the graphical capabilities of the 8810, but the work of Bill Gates at Microsoft made the company a household name, even if just in the US. In Europe, the 8810 and its licensed clones reigned supreme, and many of today’s computers in the EU can trace their lineage to the 8810 and are generally compatible with this first desktop computer.

Nixdorf Apple Computers

As for the continued history of Nixdorf Computer GmbH, in 1982, Steve Wozniac moved to Germany to begin working at Nixdorf, following a plane crash in 1981, which left him unable to continue working with the Apple II team at Apple. He had grown a bit disillusioned by the direction Apple was taking, trying to catch up with the DEC 8810 and wanting to see who had built that computer.

Wozniac proved invaluable for Nixdorf, partly for his connections in the US, but also with his knowledge. By 1985, he was also responsible for setting up a meeting between Steve Jobs and Heinz Nixdorf, which led to the acquisition of Jobs and Wozniacs shares of Apple computers, after Jobs got into a deep spat with the new Apple CEO John Sculley.

The acquisition of those shares made Nixdorf the majority shareholder of Apple, and Jobs was placed back into the CEO seat. Jobs also became something of a protege to Nixdorf, while allowing Apple to begin producing clones of the 8810 for the American market, even if this hurt Nixdorf’s cooperation with DEC.

The death of Heinz Nixdorf in March 1987 was a significant blow to his family and company. With his death, Nixdorf Computer GmbH was threatened by a loss of direction, but since the company was not publicly traded, the Nixdorf family was responsible for its direction. It was Nixdorf’s son Martin, who initially tried to take over as the CEO of Nixdorf, but found himself overwhelmed.

Finally, in 1988, almost a year after Nixdorf’s death, the Nixdorf family asked Steve Jobs to take over as the CEO of the company. While Jobs was initially dubious and had a few problems with the way Germans did things in Germany, he quickly came around to their way of doing things, something that catered to the sense of care and perfectionism Jobs had.

By 1989, Jobs had talked the Nixdorf family into a full merger between Nixdorf and Apple, and making it a publicly traded company. The Nixdorf family would retain the majority of shares in the new Nixdorf Apple Computers GmbH.

With Jobs at the helm, Nixdorf Apple would continue to innovate in the area of computers, driven by Jobs’ vision of things. This would include the first tablet computer and smartphones in 2002 and 2004 respectively.