NCR Safari 3115

February 27, 2020


Recently, on ebay, I stumbled upon another NCR mobile, pen-based computer. In contrast to my beloved 3125s, I have never heard of it.

It looks like the older, smaller, much uglier brother to the 3125, but it is not. It is the much rarer, smaller, much uglier, younger brother of the 3125: the Safari 3115.

“Safari” already hints to the time it was released, because Safari is the family name of the AT&T mobile computers. NCR was taken over by AT&T in 1991/92, so this release must be later than that.

Now, the model designation 3115 puts it in the NCR 3000 family, below the 3125, but the “3000 family” was always a marketing lie as the computers in it differ a lot architecture-wise and OS-wise.

The operating system of the 3115 was Windows 3.1 with pen support. The RAM seems to consist of a DRAM card that is accessible via the interface bay. Mine is a Samsung 4 MB model. It is labelled “ICMC V4.1” which is the same as PCMCIA 2.0.

There is an “unofficial NCR (Safari) 3115 support page” which holds most of the information on this machine, but it seems to updated last in 2000…

It weighs 1700g, so it is even 200g heavier than the A4-sized 3125… And although the outer area of the 3115 is smaller than the one of the 3125, the outline volume is probably about the same or even larger!

The Computer History Museum even has a docking station called “CommStation” for the 3115.

I did not think that this machine even deserves more research, because, boy, it is ugly, and because there were a lot of Windows-based pen tablets out there after 1991.

However, I opened the case to see what’s inside because some things (like the concrete CPU model) are not even known. And I have to say, the inside is more advanced than what I thought it it would be.


What I was surprised first was the fact that the computer seems to be at least splash-proof. There is a thick seal in between the case halves, and most of the interfaces are put behind rubber plugs. There is not even an electric interface connecting the machine to the docking station, but an array of infrared LEDs.

The second suprise was that most of the weight seems to be contained in the display which is secured in a metal case. The PCB is not very large, the case not that heavy.

The final suprise was the “harddisk”. It is not a harddisk, it is basically the second generation prosumer-grade SSD in form of a SunDisk(sic!) SDI-20 20 MB SSD with a date of 1992-1993. The first generation was developed by SunDisk (now SanDisk) for the original IBM ThinkPad in 1991. It had a capacity of 20 MB and costed $1000. Obviously, an SSD is much more suited for a rugged pen-based computer than a rotating harddisk.

Other things I found inside were: an Intel 80486SX CPU and 4 MB of soldered RAM.

So, here is my final verdict: This was an attempt to created a rugged, small, pen-based Windows tablet. Unfortunately, it is way too heavy and ugly. Although I don’t know the original price, it probably have not been cheap. It is more interesting that what meets the eye and it is a very, very rare thing.

Technical Data

Manufacturer: NCR
Model: Safari 3115
CPU: Intel 80486SX@25 MHz
RAM: 4MB (8 MB max)
HDD: 20-40MB
Weight: 1700 grams
External dimensions: 23cm*23cm*7cm
Pen: Cordless 1 button digitizer pen made by CalComp Inc, transmits on R/F ranges 0.0576Mhz and 0.0614Mhz, runs on 4(E 393 buttoncell batteries)
Display: 6.25″ Backlit Monochrome VGA
OS: MS-DOS 5.0/Windows 3.1 with pen support
Interfaces: 1 RS232C DB9 serial port, 1 Centronics 25-pin Parallel port, 1 PS/2 Keyboard port, 2 PCMCIA-II expansion slots, 1 PCMCIA-1 memory card slot, 1 Infrared communications/docking port, 1 external power/charging connector
Battery: 9.6V (NIMH) 1200mAh, 1.2A, good for 4hrs per charge?
Released: 1993
Initial price: ???


Sony NEWS Portable Workstation NWS-1250

February 27, 2020


We do not see many vintage Japanese workstations in the West. We also do not see many portable workstations (of any make). So, what we really rarely see, are Japanese portable workstations. I feel quite fortunate to have recently acquired one, even if it is not working as of right now. But, once more, let’s start at the beginning.

Sony is, by-and-large, a consumer electronics company and is well-known for their MSX offerings back in the day and their VAIO line of PCs until quite recently. Not so well known is the fact that Sony produced classic workstations from 1987 to 1995. By “classic” I mean they started off by Motorola 68k-family CPUs and ending in RISC CPUs, and they used Unix as their operating system.

The story of why Sony was producing a class of computers that did not really fit their usual lineup goes like this: The manager of a newly established computer project wanted to create a new, cheap office computer, but his engineers wanted instead to do a computer that would replace the VAX 11/780 that was hard to get computer time on, but which they used in earlier projects. The manager settled with the wishes of his engineers and out came a machine that gave each engineer his own substantial, graphical compute power. Now, the concept of workstations is well known in 1986, but Sony competes in the market via the cheaper price and the better price/performance ratio.

The first generation of Sony workstations (1987-88) had mainly dual 68020 CPUs in them and often a Motorola 68881. The second generation (1988-90) used a single or dual 68030 CPUs and a 68882 mathematical co-processor. The third generation (1989-95) used MIPS CPUs, first a R3000, then an R4000, R4400, R4600, R4700, and the final model in 1995 used a R10000.

The operating system, NEWS-OS, was first a BSD variant, in later iterations an System V derivate.

In the long list of 60 models Sony produces, there are also two models (with two variants each) of portable workstations. Both models use the same case and look the same.

The earlier models, the NWS-1230 and NWS-1250, are second generation and have a single 68030 CPU and a 68882 co-processor. They are released in 1990 at a price of 1,250,00 and 1,550,000 Yen, respectively. The slightly larger model, NWS-1250 is the one I got.

“Portable workstation” here does not mean laptop computer, although it is certainly possible to place it on the lap. But without a battery, a weight of about 8 Kg, and measuring 35x42x10cm this is not exactly something you pack in your bag. The machine contains a black-and-white LCD, a keyboard, a 3.5″ floppy disk drive, and a (SCSI) harddisk.

How rare were portable Unix workstations at that time? Let’s have a look at the competition in 1990. There are, of course, many portable PCs available, although typically not of a workstation class. The only other portable Unix workstation that I found before 1990 are the 1985 HP IntegralPC (with a 68000 CPU and Unix in ROM) and the 1989 Opus Portable Workstation (with a 88000 CPU) for $14,000. The next models that I found are the 1992 Tadpole SparcBook 1 and the 1994 Sun SparcStation Voyager. All in all, it seems to me that in 1990 there weren’t many portable Unix workstations around (if you do not count PCs with Unix). And also: why should there be battery-less portable Workstations? The main use is probably showing potential customers your Unix programs, and that is a real niche market.

However, the design of these portables is still quite beautiful and thought-through. When you close the lid with the LCD, it brings your keyboard in a horizontal direction and basically seals it off so no dust can enter in transport. When you open the lid, the keyboard swings down a few degrees, giving you a nice typing angle.

Apart from the design, there is nothing too special about the hardware except the audio capabilities. This machine has dedicated headphone and line in, and mic sockets right at the side of the keyboard. There is also an “MIC ATT” switch (0 db and -20db) and one that switches between mono and stereo. There is dedicated audio hardware onboard, but I do not know the concrete characteristics of that.

Now, my machine is not functional because it lacks a harddrive. These things used SCSI drives and were purportedly picky with which model they chosed to collaborate (there is one guy who says that the kernel has a list of allowed harddisk models). My model had a 240 MB harddisk, probably a Hitachi DK312C-25. Let’s see whether I bring it to life, one day.

Technical data

Manufacturer: Sony
Model: NWS-1250
CPU: Motorola 68030@25 MHz
FPU: Motorola 68882
RAM: 8 MB, extendable to 12 MB
Harddisk: 240 MB
Floppy Disk Drive: 3.5″, 1.44 MB
Display: 11″ b&w LCD, backlit
Graphics: 1120×780
Interfaces: mouse, audio (phones, line in, mic in), SCSI, Ethernet (AUI), serial (DB9), parallel (proprietary), SCSI (SCA?)
Released: 21.07.1990
Initial price: 1,550,00 Yen (or 25,000 DM)


A few remarks on the Canon Cat

February 23, 2020

I’m currently writing a German article for the club journal (“LOAD“) of the VzEkC (basically the national German vintage computer club).

Here are some remarks that did not fit into my article.

Cat Manuals

There is a whole load of manuals for the Cat available electronically. Simply visit for the

  • How-to Guide
  • Reference Guide
  • Online Help (the files from the Cat ROM)
  • Tutorial (the file from the Cat ROM)
  • Quick Reference Card
  • Hardware Schematics
  • Workshop Manual
  • Technical Documentation for the Canon Cat Editor
  • tForth Manual

LEAP in the 21st century

One of the main features of the Cat was Jef Raskin’s LEAP technology. It is basically a fast search-in-the-entire-document-space (in the memory of the Cat) plus moving the cursor to the found location. If you can remember a name, a location, or a phrase, you can find the corresponding location across all documents on the Cat.

This would be a useful function even (or especially) today if you could do it over all things you have ever typed in on all machines that you have ever used. It would require some technology to track all things typed in by a user, but nowadays we have enough compute power to search in that type of data amount quickly.

Finally, although there is a lot of things you can find about the Cat electronically, I received with my machine some things that I did not find in the Internet.

Canon Cat prospectus

It seems to me that this must be the standard marketing material.








Obviously, there was also a little bit oft software for the Cat. I do not know whether these would be sold separately, whether the shop simply copied them to customers, or what their status was. For the CATFILE program a “Note to dealers” basically says that this should be given out to customers so they “will appreciate this usefull addition to their Cats, and we hope they will be encouraged to tell their friends and business associates about this recent upgrade to their Cat systems”. Does not sound as if the Cat sells itself easily, does it?

CATFORM is a Cat program that allows to manage forms on a Cat and to print them out. You can create, store, and fill out forms with this program. I have a printed manual for that. This manual is also available electronically under:

CATFILE allows yout to manage name and address lists on the cat and to use them for mail merge. I have a printed manual for that.

CATFILE Utility Disk
Now this program is not directed towards the users, but towards the dealers. It “is used to create and customize the Catfile Application Software disks for your customers use. I have a photocopied manual for that.

“Secretarial Workstation” demo disk and marketing material
This package was meant for dealers to demonstrate the “secretarial workstation” potential to potential customers. To that end there was a demonstration disk and a short printed manual that explained how to use the disk.

Family Tree v2.1
I do not know what this app really is. Obviously, it is some form of tool to record data for a Family Tree, but I do not know whether this was a commercial program or whether the last owner did program this himself.

BTW, if everythings works out, I will show off my Canon Cat (and the software) at the Classic Computing 2020 exhibition in Thionville, France on the 26th and 27th of September 2020.

Western Digital Pascal MicroEngine, an Update

February 16, 2020

Yesterday, I received a 2nd MicroEngine board, unfortunately in a worse state than my first one. Also, probably an earlier patch level as there is less patch cableing on the board. Nevertheless, nice to have.

Also, if you look for images for boot floppy disks, you can find them under

Umtech VideoBrain Family Computer

January 10, 2020

Recently, while searching for another computer, I found a computer in a box in another box that I obviously bought some years ago, but then completely forgot about it. What was even funnier is that I saw this exact model on ebay some days ago and actually thought about bidding for it (which I didn’t). This computer finally sold on ebay for over €850 or $960 even without power supply or joysticks. I do not remember, but I am pretty sure I did not pay that type of money whenever it was that I purchased it.

Okok, to the point: Let me introduce you to the VideoBrain, sometimes also called the VideoBrain Family Computer, the sole model produced by a company called Umtech from 1977 on.

Very unusual for a home computer was the full travel, 36-key keyboard it featured. However, it is said to be poorly designed and difficult to use. Certainly, the choice of functions on it looks very weird.

The VideoBrain had no built-in computer language. However, there was one language available as a cartridge from December 1978. If you would had to guess which language one should offer to home consumers, only a few people would come up with the choice of VideoBrain: APL/S. APL is a programming language which (according to its Wikipedia page) is influenced by “mathematical notation” and influenced itself such crowd pleasers as “MATLAB” and the “Wolfram Language” :-).

Regarding the VideoBrain and APL/S there is a magnificient audio recording from the Third West Coast Computer Faire conference in 1978. This recording was made during the presentations of the Conference. Thir first and second presentation are:

  • Ted Haynes: Videobrain and the APL/S Language
  • Robert G. Brown: An Introduction to APL/S: A Modern Computation Language for Personal Computing

and you can listen to them in very good quality using the last link in the references.

And, you heard it here first, the $150 APL/S cartridge actually had not only ROM, but also more RAM (if you listen to the presentations from the last paragraph). This is also hardly surprising as 1 kB of RAM is hardly enough for a (later) Sinclair ZX81 with its efficient, token-based program representation, let alone APL, a language that can handle entire number arrays with a single operator. <update>[Brown78] states that the cartridge had 13 kB of ROM and 1(!) more kB of RAM.</update>

When you listen to the presentation of Robert G. Brown, you can also hear his rationale of offering APL as opposed to Basic, namely the higher productivity and better degree of programming structures. These are, of course, honorable reasons for a computer scientist, but seem a little bit odd for an entry-level computer with hardly enough memory to actually have program readability problems :-).

The above presentations are also contained as text articles in the Proceedings of the Third West Coast Computer Faire conference. But if you are interested in the APL dialect, also Byte had an article by Robert G. Brown, the author of APL/S in the December of 1979 issue, and this issue is online (see references below).

The main differences of APL/S to APL are:

  • APL/S is a subset of APL
  • all these pesky special characters the original APL needs as operators (like the arrow and the star-im-a-circle) are replaced by ASCII strings
  • arrays in APL/S are restricted to one dimension and subscript expressions must evaluate to scalars
  • APL/S adds control structures like IF and WHILE

APL/S uses a two-part user interface. In the lower half the user can enter and execute code. The upper half is reserved for bar charts. For variable names, only the first four characters are used.

The VideoBrain was the first home computer system where the software was available as cartridges. These could contain of up to 12 kB of ROM. Fewer than 25 software titles were ever markteted for the VideoBrain.

The used CPU is quite old and exotic. It is a Fairchild F8, which consists of two chips (the CPU and the “Program Storage Unit”. This sounds awkward, but was actually a technical achievement in 1975, at a time when earlier CPU designs distributed the needed functionality over a larger number of chips (sometimes 7 or more chips). Later on, CPUs assembled all functions in one chip. As a result, the F8 was quite economic, which, according to [CPU Museum], made it in 1977 to “the world´s leading microprocessor in terms of CPU sales”. However, as we know, the number of computers exploded only after 1977, and these computers used other CPUs. Therefore, it is not surprising, that there aren’t that many F8-based computers. There is the VideoBrain, and there is Fairchild’s own console, the Channel F Video Entertainment System from 1976.

Due to the high cost of RAM at that time, the machine came only with 1 kB. However, it had 4 kB of ROM, providing four built-in programs: a simple text editor, a clock, a count down timer, and a color bar generator.

The basic computer itself does ntot have any possibility to save data, you had to buy the “Expander 1” if you wanted to have cassette tape recorder interfaces (and two RS232 interfaces). The “Expander 2” was a 300 baud modem.

Graphics on this machine seems to be complicated. [SeanRiddle] says “This document describes the VideoBrain grapics hardware. It is a sprite engine, capable of displaying 16 sprites simultaneously. The control registers are documented pretty thoroughly. The sprites are monochromatic, but each can be a different color. There is one bit each for RGB, and 2 bits of intensity info (but maybe only 2 intensity levels are usable). The registers allow for a sprite up to 248×256 pixels, positioned on a grid 256×512 pixels in size. Sprites can be displayed at twice their horizontal or vertical size. There are 2 “display lists” for setting the y position and drawing priority of the sprites. There is also a mode called “xcopy” that replicates the first byte of a sprite horizontally.”

The computer was not widely available, but was sold for a short time by Macy’s department store. As you can imagine, the VideoBrain was not a large success, and it vanished from the market after 3 years.

The Wikipedia article on the VideoBrain talks about the fact that the VideoBrain had no real defined target audience and therefore could not satisfy anyones needs (in contrast to the Apple II), and that’s certainly true. From my point of view, the VideoBrain tried to be both a games console (small memory, cartridge slot, 4 joystick interfaces, no cassette interface) and a computer (full keyboard, programming language available from 1978, computer interfaces available in expansion module), but did both things badly.

Technical Data

  • Manufacturer: Umtech
  • Model: VideoBrain Family Computer
  • CPU: Fairchild Semiconductor F8 @ 1.79 MHz
  • RAM: 1 kB
  • ROM: 4 kB
  • Graphics: 16 colors, sprite engine (see above)
  • Interfaces: 4(!) joystick ports, TV RF connector, cartridge, expansion port
  • Released: 1977
  • Initial price: $500 (basic device)


Gepard – Part Seven: The Gepard Company

January 6, 2020

As with many startup companies, the history of the Gepard company is very intense, but unlike many unsuccessful startups, it is both rather short *and* produced tangible results in time.

So, let’s start at the beginning.

It is the year 1983. The computers in the consumer market are mainly 8-bit computers. It is already clear at this point that the Next Big Thing will be 16-bit computers, but in Germany you can buy 16-bit computers only as expensive, professional computers. Examples for these models are:

  • the HP 9816 (starting at $3895 or about DM 14,000)
  • the Tandy TRS 80 Model 16 (admittedly more a semi-professional machine) at about DM 17,500
  • the Fortune 32:16 at DM 25,000

In the US, there is the “Dtack Grounded” single board computer sporting a Motorola 68000 CPU at the maximally possible 12 MHz. It uses SRAM in order to use this speed, but the board does not offer too much more. The people behind this computer also publish a newsletter of the same name which is informing its potential customership about the current developments regarding this board and the industry.

Having this in mind, a group of friends aims at founding a computer company that targets a computer these group would like to buy themselves. The assumption behind this strategy is that such a computer should meet the needs of a market of similar minded people. The requirements of this computers would be:

  • a Motorola 68000 CPU
  • modular design for maximal flexibility
  • initially designed as an add-on for an Apple II computer because:
    • the friends are Apple fans
    • this approach has the advantage of not needing to develop own UI and storage hardware from the start
    • this allows for a cheaper price for the first product

This company, the “Gepard Computer GmbH & Co. KG” is founded in Oldenburg, Germany in August 1984. The initial founders are:

  • Thomas Schumann
  • Herman Spille
  • Klaus Onnen

Later on, two more founders join:

  • Meinolf Schneider (nowadays: Amekudzi)
  • Bernd Heyer

In order to finance the company, each founder has to contribute DM 40,000. More money is contributed by an investor. The inital round of financing sums up to DM 410,000. A second round adds additional DM 180,000.

In the spirit of Dtack Grounded, also Gepard published a periodical newsletter on paper even before the actual company was started. It informed potential customers and interested parties about the current and upcoming state of Gepard developments, it provided information in different, related topics, and it dissed the competion. The newsletter was completely free. It was published from October 1983 to March 1986 in 12 issues of typically 12 A5 pages in an edition of 500 copies each. The newsletter lead to several contacts, not least it was read by a contact of the later investor of Gepard.

Development of the hardware starts already in 1983. The founders and some first employees design three cards (CPU, RAM, Apple II Interface) for the Gepard, a backplane, an Apple II card (which connects to the Interface card), the Modula-2 compiler, the operating system (GDOS), and the software that runs on the Apple II. They have the PCBs then produced locally (in Oldenburg), buy components, and populate and solder the PCBs themselves. A first run of 30 sets is produced and in 1984 the first sets can be sold to the people who had registered via the newsletter.


Marketing-wise, Gepard was not very active, but tried to get the most out of its limited effort. The main marketing channel was the Apple User Group Europe (AUGE). So it distributed the newsletter generously via AUGE channels. In the newsletter readers could register an option on a Gepard for free. 30 options were registered and 3/4 of the options were resulting in purchases. AUGE was divided in regional chapters, and in many of those chapters there was one Gepard owner, so there was always one to present a Gepard or who could be asked for opinions. As a consequence, and because of a 10% commission Gepards are mainly sold by word of mouth and by existing customers. There was also an article in the AUGE magazine (8/84) on the Gepard. Articles on the Gepard are very rare, not least because at that time, journals prefer promoting their own computer projects than reporting on a new computer with a very small customerbase. Nevertheless, Gepard also runs (small) ads in the most important computer journals. Finally, Gepard is represented with (probably very small) booths both at the Hannover Messe 1986 and at the Orgatechnik 1986.

In 1984, Apple releases the first Macintosh, also a 68000-based computer. Fortunately for Gepard, although the Mac has a mouse and a GUI, and is sold complete in a neat, compact package, it is more expensive (at DM 10,000), and it is (at 8 MHz) slower than the Gepard. It is also by and large not expandable. As also the typical customer of a Mac and a Gepard differ substantially, the Mac is no threat to the Gepard.

Until 1985 Gepard had 70% private and 30% professional customers. In the fall of 1985 Atari starts selling a new model, the Atari ST for an unrivaled low price (~DM 3,000 including floppy disk drive, monitor, mouse, and GUI) and basically ends sales of Gepards to private customers. The Gepard company does not know how to react to this competition, has not finalized all of its developments and is out of money. Eventually, 1986, the company files for bancruptcy.

In hindsight, Thomas Schumann, the former CEO of Gepard, thinks that it would probably have been possible to save the company if the management would have had more experience, and if it would have moved its focus to professional customers, offering a Unix-like operating system. After having found that the later 68020 version of the Gepard was quite unique at least on the German market for its price, I agree.

After Gepard went bust, one of the original founders, Hermann Spille founds his own company, HS Computer, to continue the Gepard business. He continues selling hard- and software from the Gepard catalogue, but also develops and sells new hardware. The most important development activity is the finalisation of the 68020 card and the development of the second version of this card. Other hardware includes:

  • a Z80 card
  • an EPROM burner

Apart from HS Computer, there are a whole number of small projects starting in the aftermath of Gepard:

  • ports of OS-9, PC DOS 2.1/GemDos, Eumel/Elan are started or at least proposed
  • version 1.4 of GDOS is finalized by Thomas Tempelmann
  • version 2.0 of GDOS is written by Harald Hellmann
  • an “Atari ST” graphics card is developed that can be connected to an Atari monitor

But eventually, over time, the Gepard business trickles off and HS Computer is liquidated in 2009.

Ok, so, here you have it. My collected findings about the German Gepard computer. I hope you liked it. It took me a long time to collect, analyse, and write, but it was fun, it was new, and it was basically never documented before (especially in English).

Gepard – Part Six: The Software

January 1, 2020

Let’s talk about the software in the order in which it was developed. We therefore start at the beginning (in Gepard terms): the Modula-2 compiler.


One cannot talk about the Gepard story without mentioning its Modula-2. Following the Zeitgeist (and Apple) not only should the operating system of the Gepard be written in Modula-2, but this computer language was also foreseen as the system language, the main application language, and the shell/batch language. This meant e.g. that all hardware was accessible to Modula-2 languages, by providing corresponding libraries.

The Modula-2 compiler was initially written by Jürgen Müller (who did not have much experience in this matter) directly in Assembler. The development started as a cross-assembler project on an Apple II. As soon as enough developer systems were produced, the development was moved to a Gepard itself, and as soon as no Apple II (or C64) was needed anymore, everything was done on the Gepard itself. As a result of being developed in Assembler, the compiler was very fast at compiling code (one Gepard manager believe it to be the fastest M2 compiler at that time in the world). As a result of the main programmer being not that experienced in that area the compiled code was not that fast (although the comparison of the Gepard Modula-2 to any Apple II compiler went of course very well for the Gepard as proven by the original comparison from 1985(?) below).


source: 68000=news Nr. 6

Users of the Gepard Modula-2 also often mentioned the completeness of the compiler positively. Being the main system language, it even featured an inline assembler.

Jürgen Müller received a royalty of 22.50 DM per Gepard for the compiler.

Later on, the Gepard M2 was ported to the Atari ST under the name “Megamax Modula-2” by Thomas Tempelmann. It was then distributed by “Application Systems Heidelberg”, which was funded later on by another former employee of Gepard. The initial version was mainly a pure port of the original Gepard version, later issues had improved features. The lower execution speed of the compiled code was also mentioned in comparative reviews of different Atari ST Modula-2 compilers.

However, the Modula-2 compiler was certainly the main reason to buy a Gepard before OS/9, because there was simply no other language you could use (except Assembler) nor any standard binary format of a program that could be executed.

The (original) Operating System: GDOS

The focus on a programming supporting software system continues on the User Interface side: it is a clone of the UCSD Pascal UI. It is all text, no mouse pointer nor windows in sight. All you have are a fixed set of commands that you reach by typing a single key. Some of the commands are displayed in the first line of the screen. If you branch into one command (which is an application on the top level), you have another set of sub command, some of which are displayed on the top.

Some of thes applications on the UI level are: Editor, Compiler, Monitor, Filer (a program that can do things with files).


This is a strange version “1.4cv”

One could argue that it would have been better to use a standard Operating System, but apart from Unix (which basically required a harddisk) there was not really one. The was CP/M-68k, but this was hardly standard and would have offered only a few advantages while being costly.

However, the lack of Unix was probably *the* problem that prevented the success of Gepards in the professional area after the Atari St and the Amiga had killed the private market for the Gepard. If it would have offered Unix, a Gepard would have been probably a cheaper alternative to “professional” 68000 computers, especially after having finished the development of the 68020 cards.

During the time of the Gepard company, GDOS was developed up to version 1.3*. Version 1.4 was finished after the Gepard company went bust. It was developed by Thomas Tempelmann, who tried to convince Gepard users to pay him some money for the version because the development was no longer financed by Gepard. It is not completely clear to me whether 1.4 was the last “official” version of GDOS because I have a version 1.5 from the floppy disk remainders owned by Jürgen Müller, who was obviously part of the special interest group GDOS maintenance who is mentioned on the splash screen of version 1.5.

A much more capable version of GDOS was always in the planning, even at the time of the Gepard company. A corresponding, much improved version of GDOS was later on published as “GDOS 2.0″ (and later on called OS/Science”) offering e.g. multitasking, different fonts, etc. However, as far as I can tell, this was never delivered with later Gepards, but needed to be bought as a separate software for DM 400.

Application Software

There wasn’t much application software as far as I can tell.

There was Gepard’s own editor (GepStar, a WordStar-like program) and Gepard’s own spread sheet program, GepCalc. Initially sold separately, it was later on bundled with GDOS 2.0.

There was Gepard’s own PCB layout program, PlaTool, which Gepard used itself to design new cards.

Then, there was a small catalogue of little tools like GepTerm, a terminal program, a disk with utilities, one with test tools, disks with Modula-2 libraries for mice and graphic tablets, the GDC graphics, and the graphics of the 80-characters-card. There was a small sampler program, and two disks with samples.

For GDOS, GepStar, GepCalc, and the Boot ROM images, you could also get the source code.

The other Operating System: OS-9

OS-9 is a Unix-like multi-user, multi-tasking operating system originally developed for the Motorola 6809 processor by a company called Microware Systems Corporation. It was originally developed from 1979 as the operating system for the BASIC09 project fundedy by Motorola. At that time it was used mainly for TRS-80 Color Computers, and Dragon computers (requiring a floppy disk drive). In 1983, OS-9 was ported to the Motorola 68000 CPU.

In contrast to most operating systems, OS-9 was sold mainly as a 3rd-party-product that needed to be purchased by the users. Therefore, also the Gepard version of OS-9 was sold for about DM 750 from about the end of 1986 onwards. At Gepard times it had a small but active community including an own newsletter. On the Gepard, OS-9 uses its own file system and cannot be copied by GDOS tools.

You can still get OS-9 for many computers today.

Even if you cannot copy OS-9 using GDOS, you can copy it using OS-9. First, boot OS-9. As the user name, use an empty string, i.e. just hit ENTER. Now, you can format a new floppy disk with format /d0 (if you want to use the first drive) or format /d1 (for the second drive). Using the backup command, you can copy disks. For a cheat sheet of OS-9 commands, look onto the last page of this file.

Software, that did not really make it

CP/M 68k

This actually appears in the 1986 price list including a C compiler for DM 1225. I know someone who claims that he actually tried it. However, I do not know anybody who claims to have actually used it. Technically, it would not have been impossible to actually port it. Most probably, the price was basically determined by the license fee and it offered not that much of a benefit compared to GDOS.

PC DOS 2.1/GemDos

In 1986, in the aftermath of Gepard Computer GmbH going bust, someone asked for subscriptions to a project that wanted to port GemDos to the Gepard. It would include PC DOS 2.1 and the GEM Desktop and would cost DM 499.- If 100 users would subscribe, they would buy the license from Digital Research and start porting the software. I never read anything about this project again, so probably the 100 subscriptions never happened. Also, it seems a little bit like a stretch to port PC DOS to the 68000.


Elan (Education Language) is a programming language that was developed in 1976 in Germany. The Operating System to run it was called Eumel. Like the GemDos porting project, this was an attempt by an individual to open the Gepard to other software worlds. Like GDOS was written in Modula-2, Eumel was mostly written in Elan. The price for a multi-user version would have been DM 750.-, but like the GemDos port, this project seems to have never materialised.

Gepard – Part Five e): The Hardware: GDC Graphics

December 15, 2019

The Gepard 80-characters card is able to display also some monochrome graphics, but if you wanted color you had to buy the GDC graphics solution. This solution is rather unique in that it is a modular graphics subsystem centered around the NEC 7220 chip, which is a marvel on its own. But let’s start at the beginning.

A GDC graphics subsystem consists of one Video card and 1 to 8(!) GDC cards. The GDC cards are “daughter-boarded” (or to another GDC card) to the Video card, and only the Video card connects to the Gepard bus, making it a rather bulky PCB block. You can see this on my GDC block consisting of 3 GDC cards on top of the Video card.


Each Video card cost around 300€ and each GDC card around 500€. This means the maximum configuration cost 4300€, that’s more than an entire (text-capable) Gepard computer…

The Video card has pin header connections for:

  • a monochrome Composite monitor signal (the same as from the 80-characters card)
  • an analog RGB monitor

The GDC card each have a NEC 7220 graphics processor and 128 kB RAM.

If there are more than one GDC card, the cards can share the workload by being responsible only for a part of the color information each.

For example, if you have 8 bits of color information per pixel, you can share it amongst 1, 2, 4 or 8 GDC cards. As a positive side effect you can have higher resolutions as each card has to work less on a single pixel in the restriction imposed by the time a single frame needs to be created (which is given by the time the CRT tube needs to display the single frame).

Depending on the RAM need of a single frame, the number of GDC cards, and some other restrictions, you can have more than a single frame (up to 16) drawn in parallel (then simply switching to another part of the RAM).

You can have up to 256 colors out of a palette of either 2^12 or 2^18 colors (the manual and the price list differ in this point).

The highest achievable resolution seems to be 1024×102 pixels.

It is not possible for me to list here all possible resolutions, color depths, and number of parallel screens as this depends on:

  • the number of GDC cards
  • the type of monitor interface used
  • the jumper configuration on the GDC cards

Also, not all combinations are possible.

The NEC 7220 chip is probably the first integrated “GPU” chip. It was released in 1982 and had the following features:

  • DMA-capable
  • own memory control
  • own command set: lines, circles, arcs, rectangles, characters
  • can address up to 512 kb RAM
  • max. resoution: 2048×2048 pixels
  • can use a lightpen directly

Its importance can also be seen from the fact that Intel licensed the design and produced it under the Intel 82720 designation. Which in turn triggered the Soviet block U82720 clone.

This graphics architecture has the advantage of being

  • very scalable
  • easy to incrementally extend

but also to be

  • very expensive (in 1985, my 3-GDC-card block is more expensive than an entire Atari ST system a year later on)

Therefore, GDC graphics seem to be rather scarce. Out of the 6 Gepards whose configuration I know only mine has GDC graphics, the rest has 80-characters cards.

Also, there are hardly any other computer models based on NEC 7220 graphics that have more than one such chip. The DEC Rainbow 100 (graphics option), Epson QX-10, QX-16, Tulip System-1, EC 1834, A7150, as well as the Number Nine Revolution 512 family of graphics cards all had one NEC 7220 (or equivalent) chip. Only the NEC PC-9801 and the NEC APC family had two, supported by 256 kB RAM.

In later years, GDC cards had 512 kB RAM each.

Gepard – Part Five d): The Hardware, Misc. Topics

December 2, 2019

To conclude the hardware parts, here some short remarks on various
“hardware”-related topics.


Serial Numbers
Easy topic. There weren’t any. Neither on the case nor on the cards. However, some cards have a paper sticker with a date on it.

Card Versions
My above remarks on the different cards might have give you the impression that there was only one version of every card. Well, there wasn’t. Some cards have even version numbers on the PCB. See the above picture. CPU card V1.3. Sticker with a “18.11.(19)85” date. Hmm, I could have mentioned that Gepard designed the PCBs (later on with a program on a Gepard), a company in Oldenburg, Germany (where Gepard was located) produced the (naked) Gepards, and Gepard itself populated them. Quite short turnaround time I would assume, but neither the fastest nor the cheapest way. Quality, local production, though. These were the times where assembling stuff in East Asia was not the fast, cheap, (sometimes) quality way to go for every company.


The Manual
No Gepard system was complete without the (up to) 650-page custom-made A5 ring binder. “It was not complete” is not a phrase. As the operating system, the hardware, and the system programs were all completely proprietary, this was the only document that described them. The Modula-2 compiler was near-standard, but not completely, so you also needed a description of the details (e.g. in the 100 page annex). For good measure, it also contained a 144-page introduction into Modula-2 (which made sense as this was exactly the audience that would buy a Gepard). A description of the card hardware (incl. block diagram, population plan, pins and jumpers as well as the occasional Assembler code) allowed the user to use and modify the hardware. The manual was completed by a description of GDOS (the operating system), the “Filer”, and the “Monitor” programs as well as a 100-page description of the Motorola 68000 CPU. Everything you need in a pre-Internet age.

Oh, yes, all documentation is German only 🙂 (well, the nouns are often English as is usual in technical German). You can find two version of the manual in the electronic resources post.

Gepard – Part Five c): The Hardware, Phase 3

November 25, 2019

After the Gepard company went bust, one of the founders, Hermann Spille, founded HS Computer and continued providing Gepard hard- and software and even developing further cards. In this last phase, the 68020 cards were finished and put on the market. In this phase, most of the cards from the 2nd phase were still available. HS Computer was liquidated in 2009. Afterwards, no new Gepard component development or distribution seems to have taken place.

Third Phase: 1987: the 32-bit era


68020 cards

There were two variants of this card. In April 1987 the first version appears. It sports a Motorola 68020 with 12.5 MHz, offers a socket for a 68881 FPU, and 128 kB SRAM. The problem with the Gepard architecture and 32 bit CPUs is that the used proprietary bus is a 16 bit bus providing mainly the signals of the 16 bit 68000 CPU. A 32 bit CPU needs a 32 bit bus if you want to connect CPU and RAM at full speed. In a backplane-based system like the Gepard CPU and RAM reside on different cards. Now, unfortunately, you cannot change the bus in a backplane-based system because you would have to change a lot of hardware and still have the problem of backwards compatibility with the old 16 bit cards. So, the typical solution, also used here, is to put (some) RAM on the CPU card (where you can have 32 bit connections), and use the system bus for the rest of the system. The first version provides only a small amount of on-CPU-card RAM (128 kB) and tries to mitigate this small amount by using a faster RAM variant, namely SRAM. The bulk of the RAM is still residing in the separate RAM cards, connected by only 16 bits, but by caching memory in the SRAM now not every RAM access needs to pass the 16 bit bottleneck. Although not backplane-based system, the same problem and solution occurred also for the Amiga 3000 and Atari TT. They both sport the legacy 16 bit architecture of the Amiga and Atari ST, respectively. This means that both new systems basically contain an old system, but now have a 32 bit CPU. That’s why in both systems you have now two different types of RAM: old, 16 bit-connected RAM, and new 32 bit “Fast” RAM connected directly to the CPU.

The second version (in DEcember of 1987) of this card uses 16.5 MHz, has also a 68881 socket, and sports now an own 32 bit socket for a daughter memory card. At the same time, a 4 MB (D)RAM card is offered for this socket. This allows now to access all memory with 32 bit.

In the research of the Gepard history it occurred to me that probably from 1987 the 68020 Gepards were the cheapest 32 bit systems in Germany until the appearance of the Amiga 3000 and the Atari TT in 1990.


Z80 card (“HD 64180 – Karte”)

This card contains a Z80-compatible HD 64180 processor that runs at 8 MHz. It provides also 256 kB of RAM. Additionally, it features a Centronics printer port and two serial ports. It is delivered with an LDOS called operating system that is compatible with CPM 80 V3.0.


EPROM Burner

This card allows to program two EPROMs at the same time. This card has a cable that connects to a small external box with two EPROM sockets. The needed software is delivered with the card.


“Atari” Graphics card
This graphics card was developed by a company that two of the other founders of Gepard founded. It offeres basically the monochrome graphics capabilities of the Atari ST having a resolution of 640*400 pixels black and white and have an interface where the 70 Hz Atari monitor (e.g. the SM 124) can be connected to. This card can also be used for generating text output instead of the 80-characters card.