This is Beagle-Ears -> Lars -> Professional -> Computer History -> My History

Computer History

Computers I Have Worked With

If you are interested in this stuff, you should read the newsgroup alt.folklore.computers.

1960/1969: RC-GIER

The GIER computer was built by the long-departed Danish computer manufacturer RegneCentralen. In those days, computers were large and hand-built in small numbers for lots of money. The GIER was considered very successful. I think they built about 40 of them over a 6-year period. I did not actually get my hands on these until 1970, when I enrolled at University of Copenhagen, where the Math department had two in their basement, and the Niels Bohr Institute of Physics had another one.

The GIER machines were compact and handsome for their time. The cabinet was about 200 cm high (6.5 feet), 220 cm (just over 7 feet) long and about 60 cm (2 feet) deep. The two broad sides each had 3 sliding teakwood panels, the ends were a light gray-brown enamel, and off to one side was a desk console with a high-speed papertape reader (2000 cps, optical) to the left, a slanted bakelite panel with lights, switches and a loudspeaker to the right, and an electric typewriter in the middle.

The machines had 42-bit words (40 bits of arithmetic precision and two "flag bits"). They had 1024 words of this, with an optional 4096 additional words of "buffer memory" which could not be directly addressed, but could be block copied in and out of main memory. They also had a magnetic drum with 960 tracks, each of which could hold 40 words. This adds up to about 200 kilobytes of storage. Amazingly, they packed onto that drum a small operating system, an Algol-60 compiler and a runtime library with virtual memory management (for code segments only). The machines were quite usable, and were used both for undergraduate instruction in computer programming and for many research projects including weather modeling.

1960/1970: IBM 7094

For its time, this was a supercomputer. IBM donated one to the Technical University in Copenhagen, where it served as the Campus Computing resource, and also made time available to other universities. This machine had 36 bit words, and I think 32 K words of main memory. The machine was very fast: Only about 3 microseconds per instruction, so rather than let the machine read punched cards and write to the printer, all the jobs and data were read onto magnetic tape and then the output was written to tape. The conversions between tape and paper or cards were done on an IBM1401, which was later replaced with an IBM-360/30 which could accept jobs remotely via modems.

The operating system of the 7094 was called IBSYS and lived on a tape drive. It had many compilers, including Fortran II, Fortran IV and COBOL.

(A reference manual for the very similar IBM 7090 is here - PDF, 156 pages).

1965/1970: RC-4000

The RC-4000 was the successor to the GIER, but had nothing in common with it. It was a 24-bit machine with a real-time operating system that had interrupts, memory protection and a limited amount of multi-tasking. It was used for many interesting applications, including real-time chemical process control, large databases (telephone directory call-centers) and time-sharing. The operating system was the subject of a fair number of computer science research journal articles, and many of the concepts embodied in it were absorbed into Multics and Unix. The one we used at Univeristy of Copenhagen belonged to the Chemistry department, but time-sharing terminals were found all over the science campus section, where they served other departments until the Univac 1106 arrived.

1968/1972: IBM 360

The IBM-360 family of computers ranged from the model 20 minicomputer (which typically had 24 KB of memory) to the model 91 supercomputer which was built for the North American missile defense system. Despite their differences, all these machines had the same user instruction set; on the smaller machines many of the more complex instructions were done in microcode rather than in hardware. For example, machines in the lower midrange did not have multiplier hardware, but the microcode implemented multiplications by repeated addition. It was rumored that the smallest machines did addition by repeated increments!

The machines had different operating systems. The smallest machines could not really support an operating system and were often used for specialized applications, where a program was loaded from binary punched cards at startup. The middle range used a system called DOS (not related to MS-DOS) and the higher end system was called OS/360. These were the machines that established 32 bits as the standard for computers.

The first IBM-360 I used in Copenhagen was the spooling front-end for the 7094. In 1970, the technical university installed a 360/65, later upgraded to a 360/75. When it came in, it had 1 MB of RAM (magnetic core memory in those days) and a roomful of disk drives, probably adding up to about 200 MB.

Today's S/390 mainframes are direct descendants of the IBM-360 family.

1968/1970: IBM 1130

The IBM-1130 was a mini-computer built in the shape of a desk. The ALU had 16-bit words, and it came with a 1.5 MB hard disk. The DOS had a Fortran-IV compiler, but even though the machine was similarly sized to the GIER, it was much less usable.

Eventually, the Niels Bohr Institute decided that it could be used as remote job entry terminal to the IBM computers at the technical university.

More about the IBM-1130 at Howard Shubs' web site.

1965/1970: Univac 1100 Series

The Danish universities decided to install 3 large computers at the three largest univerities, an they wisely chose to get an IBM, an UNIVAC and a Control Data. University of Copenhagen got the Univac, and it was a great system.

When the machine was installed, it was a Univac-1106 with 131 K words (of 36 bits), i.e. about 600 KB. About 18 months later, it was upgraded to an 1108 which ran twice as fast. This upgrade consisted in replacing a divide-by-two flip-flop in the system clock circuit by a jumper. We also got more memory, I think we doubled it.

This machine served an endless stream of batch jobs from both local and remote card-reader/printer stations plus about 50 interactive display terminals.

CDC 6500

PDP-11

Digital Equipment Corporation's PDP-11 family of minicomputers was extremely successful for over 25 years - for good reasons. The instruction set was elegant and flexible, the engineering design was modular in ways that not only allowed the manufacturer to custom build machines of many different price/performance levels, but also allowed users to expand them in the field later. Machines existed in all differnet sizes; towards the end of its lifespan, the range spanned from personal computers built into a terminal, to time-sharing multiuser systems capable of serving as a common computing resource for an entire university department.

Read more on my PDP-11 page.

1978/1981: VAX Family

The PDP-11 address space eventually became too small to do practical work: As the machines got less expensive, people were attacking more complex problems, requiring larger programs. So Digital Equipment built a larger machine, called VAX-11 (Virtual Address eXtensions for pdp-11). This machine was the best design I have ever worked on. The instruction set was very powerful, although some thought it was too large to be truly elegant.

Read more on my VAX page.

Sun 3

When I changed employers in 1990, one of the attractions of the new job was that instead of a text terminal connected to a central computer cluster, this company put a workstation with a bit-mapped windowed display on the desk of each engineer. In those days, that would be a Sun Microsystems Sun-3/80.

Sun Microsystems was a company built on the discovery that single-chip microprocessors had become powerful enough that one could build a general- purpose computer around one of these chips, powerful enough to tackle the same class of problems that one would have done on a mainframe or a VAX, but inexpensive enough to give one for the exclusive use of a scientist or engineer. Programmers loved them. They ran the same Unix systems that was used on many PDP-11 or VAX sites, were programmed in the "C" programming language, and on the windowed screen you could have 5 or 6 windows, each looking like a terminal and switch your keyboard back and forth between them.

My new employer built communications equipment around the same family of Motorola microprocessors, and we used the compilers and other software development tools of the Sun machines.

1984/1989: Apple 68000 MacIntosh

The MacIntosh took the ideas behind the Sun and applied them to systems more suited to consumers: Half the price, and as easy to use as possible. The MacIntosh did not "invent" or even popularize the mouse: The Sun had a mouse. What was new was the desktop metaphor as a way of organizing the workspace, and the removal of command lines: Clicking on the pictures was the ONLY way to run programs.

I loved the Macs from I first saw one, but I thought this was way too much money to spend on a toy. My wife and I bought one in 1989, when a brother-in-law who managed a computer store got a good offer for a demonstration system, which he passed along to us. Our machine was called the Mac SE; it had one MB of DRAM and a 20MB hard drive, and cost USD 2000.

Ironically, one of the things that I found attractive about it was the knowledge that programming for the Mac was very difficult, and the machine did not come with any compiler or other programming tools. Thus I was guaranteed not to be tempted into spending all my spare time playing with the computer, but would be able to enjoy it as a tool.

Sun 4 (SPARC)

The MC68000 was a nice enough CPU, but Sun Microsystems decided they could build a much more powerful system for not much more money by designing their own CPU according to the "RISC" (Reduced Instruction Set Computer) architecture fashion of the time. Sun called their RISC design SPARC (which they said stood for something Sun Processor Architecture for RISC Computing).

The idea behind RISC was to make the instructions simple enough that every instruction could complete in one cycle of the master clock. You might execute a few more instructions that way, but since the CPU logic would be simpler, you could use the circuitry space that was saved on the CPU chip to include a larger cache memory, so fewer of the instructions and data fetched from memory would have to come from the main memory. This would allow the programs to run faster.

In general, RISC computers were unfriendly to programmers that wanted to write "assembly language" where the programmer has to describe the operations in terms of individual machine instructions. But most programmers very rarely do that. So with a good C compiler, the "weird" instruction sets did not matter.

The first SPARC machines came out around 1989 and allowed Sun to build larger servers to complement their desktop machines. Within 2-3 years, a smaller, less expensive RISC CPU was replacing the MC68000 chips in the desktop workstations.

Intel x86 Family

The Intel 8088 microprocessor was chosen by IBM for their first Personal Computer in 1980. Soon many manufacturers were building very similar machines, using also the slightly faster 8086 and 80186 CPU versions. Around 1982, the PC-AT (advanced technology) came out with the 80286 CPU. After interesting video games became available (beginning with the Microsoft Flight Simulator) every manufacturer had to make their machine indistinguishable from the PC-AT in order that it would be able to run these games. Soon the intense competition among the smaller manufacturers in Taiwan drove the prices of the look-alike "PC clones" well below the price of "real" PCs, and the market really took off.

After the MacIntosh stunned the market, IBM and Microsoft worked feverishly to produce a "Windows program" for PCs, and after several very bad versions, Microsoft Windows version 3.1 finally became a good enough imitation around 1990.

Meanwhile, Intel kept producing faster and better microprocessors, still compatible with the 8088/8086/80286 series, although the later versions beginning with 80386 also had a newer memory management system that allowed them to run Unix. Beginning around 1988, Unix was ported to 80386 PCs.

Today, PCs range from small portable ("laptop") computers to large server machines with 4-way multiprocessor CPU sets and can run any of these operating systems (plus several less common):


  $Log: my_hist.htm,v $
  Revision 1.5  2003/12/29 04:25:39  lars
  *** empty log message ***

  Revision 1.4  2002/03/17 05:36:05  lars
  *** empty log message ***

  Revision 1.3  2001/10/26 13:28:02  lars
  Replaced CMC -> Beagle-Ears

  Revision 1.2  2000/08/15 05:11:08  lars
  Fix typo in URL (missing end tag)

  Revision 1.1  2000/08/15 01:25:07  lars
  Rearranged files, moving these from computer/ to comphist/