Fortune Systems

Today, I want to talk about a relatively small company that existed in the 80s. They started off in time and with a good idea, but as so many forgotten companies, they delivered too late and then lost their strategic way after the original founder and visionary left. Little is known about that company in the Internet, there is not a single website dedicated to its memory although they produced interesting computers. So, let’s start the tale…


Funnily, despite their quite unfortunate fate, the name of the company was Fortune Systems. They were US-based and started off in 1980 on the idea of producing multi-user systems for serious business purposes like word processing, accounting, and the like. The selling point would have been to be cheaper than the mini-computer-based competition by relying on a (at that time) new, relatively cheap, powerful microprocessor and an existing, relatively standard multi-user operating system. The CPU naturally was the Motorola 68000 and the operating system was Unix in the form of Microsoft’s Xenix.

Who now feels reminded of another startup of the 80s that used the power of the 68000 and Unix to surpass the competition and thinks “Sun” is on the right track. Fortune tried to be for the business market what Sun later on became for the scientific market. Although one can compare Fortune and Sun in their approach and they started at about the same time, their machines are quite different. This is because the requirements by their respective  markets are quite different.

A scientific workstation tries to provide the user sitting in front of the machine as much compute and graphics power as possible and typically put the focus on communication very early on. The price point is relatively high due to the amount of high-capacity components. The availability of commercially available software is a bonus, not a necessity. Users will write their own software anyway or use freely available packets from similar minded colleagues. Users and administrators are typically quite computer-literate, so a complex system is acceptable. The competition of workstations were typically super-computers.

In contrast, a business-oriented computer system tries to provide all needed services to all user groups inside a company for the smallest price. The audience is typically not very computer-literate, and easy administration is a clear plus. The provided services are typically either commercial software packets or services that use them. Therefore, it is important that the computer system appeals to the manufacturers of the software. This can be achieved either by a large installed system base or by using an operating system to which the manufacturer can easily port its software. All this typically leads to an architecture where the users access the system via terminals as terminals are cheaper than full-fledged computers and can be controlled better by an administrator. This architecture allows a computer of the same CPU power to serve more users than in a workstation scenario as every user only needs some compute power per time (e.g. when he or she types a letter in a word processor). The competition of Fortune were business-oriented mini-computer installations.

A sensation at COMDEX 1981

Therefore, it was quite a sensation when Fortune announced at the COMDEX 1981 a powerful  business-oriented computer at the price of only 5000$. The model, the Fortune 32:16 with a promised 128k of RAM, Unix, and a 1.5 MB floppy disk drive was one of the first 68000-based computers (although 68000 CPU cards for e.g. S100 systems existed before) and probably the first business computer based on this CPU. Would it have been available at that time, even in early 1982, at that price, it would have been really a steal. But as you can already tell, it was not available neither in 1981 nor 1982, but only in mid to late 1983. Also the price of $5000 was not the one of a usable system as the system required a harddisk (Unix, remember?), and this added some hefty amount to the overall price.

Even in May 1983, almost no 32:16 seems to have been available as even Apple’s Lisa “Marketing Binder” (a wonderful document that analyses the competition of the Lisa) speaks of the 32:16 only as hearsay: “As of December 1, 1982, Fortune is just beginning to deliver hard disk-based systems.  Fortune dealers quote a delivery date of two to five weeks on the hard disk systems, and they refuse to dicuss a delivery date for the floppy based systems.”

32:16 family

Still, what was delivered finally in 1983, was a neat system, but not the sensation it could have been in 1982. The initial 32:16 was a 6 MHz 68000 system with a 5.25” floppy drive, 256 kB of RAM, and harddisks ranging from 5 to 20 MB. The system included one 12” monochrome display, one serial interface, 5 extension slots, and 4 memory slots for a total amount of up to 2 MB of RAM. The system also included the text processing software. According to [Lisa], a  configuration of 256 kB RAM, and 5 MB harddisk was available for $8990. A 10MB version would cost $9990 ([Lisa] cites Fortune dealers saying that they strongly recommend the 10MB harddisk in order to make the system “usable”). Available as options were serial interface cards, and tape drives with a capacity of 20MB per tape. Software-wise you could by Business Basic, Cobol, Fortran, Pascal, and C. The basic machine was a 1-user system. According to [Lisa], extending it to more users would cost an additional $1700 for the second user and $1200 for any subsequent user.

In the CC-Seller copy of June 1983, in Germany, the following competing computers were available at similar or lower prices:

  • Altos ACS 8600 and 68000 series running Xenix
  • Burroughs B20 under BTOS
  • Tandy TRS-80 Model 16 (running Xenix later on)

To compare the machine to others, one also can have a look on the Dhrystone measure as there is an entry for a “Fortune 32:16 68000-6 MHz” system with a “V7+sys3+4.1BSD” operating system. This value put this machine performance-wise between an IBM XT with 8086 at 8 MHz and a PDP-11/34A. A (later) Macintosh512 with a 68000 at 7.7MHz is about 75% faster.

The PS and XP families


Already in 1983, the initial model was replaced by two new models. The lower end PS series had only 2 memory and 3 extension slots, the higher end XP series (like the original 32:16) 4 memory and 5 extension slots. All models now come with more initial RAM and the hard disks start at 10 MB. The PS series could be extended up to 3 users, the XP series up to 9 users (or terminals). Everything else stays the same, the CPU, the case, and the periperals. The only other difference is that the harddisk subsystem of the XP is faster than the one of the PS.

I recently had a more detailled look on a 32:16 PS10 (Photos will follow in a later posting). What I found was that in that PS model, probably the same PCB was used as in the XP, but only 3 of the 5 extension slots and 2 of the 4 memory slots were populated. Unsuprisingly, both memory slots were used, each one was filled by a 256 kB RAM card. From the 3 extension slots, one was used by the (text-based) display controller, one by the harddisk controller and only one was available for e.g. a multi-serial card (it had a 4-port-card in it). So if you want to connect a display (not a terminal) to the machine, you could already subtract 2 extension slots from any configuration. The PS model I examined also had a CPU that was specified up to 8MHz. I cannot tell whether it still was clocked only to 6 MHz or higher.


Also a nice feature of a 32:16 model is the keyboard. As wide as the computer case, massive, and with many special keys such as “Execute” (no, there is a separate “Enter” key”), “Help”, “CANCEL / DEL” (a “DELETE” key also exists), a “LF GL” key, 3 keys with tilde, swung brackets,a colon etc., and 16(!) function keys. The keys are not mechanically clicky, but this keyboard is probably the only one with a dedicated wheel for the volume of the click sound! Finally, there is a space for a function key template below the function keys and underneath the keyboard there is a space to store unused templates… The keyboard that I opened during the mentioned PS model examination was produced by the Digitran Company.

Reportedly ([Warnock 2004]), the 32:16 series did not have a MMU (because it was not available yet). Now, for Unix-like systems you typically need something like a MMU (unless it’s Minix). Therefore, Fortune designed their own circuit (using MSI TTL chips) that allowed at least swapping (but not paging).

Also quite interesting is their reasonably effective copy protection scheme according to [ClassicCMP2005]: “Uninstalled Fortune software on distribution media was encrypted using a key known to Fortune and to Fortune’s installation program.  When you installed software from the distribution media, the software would be decrypted and then re-encrypted using a key based on the motherboard serial number for storage on the hard disk (so you couldn’t just copy the executables from your system to some other system: installed software only ran on the system on which it had been installed); and of course the installer marked the distribution medium as “installed” so you couldn’t just go install it again somewhere else.”

Price per user

The significant value for a Fortune customer was always the price of the system per user. Therefore, let’s have a look at them. The following calculations are based on the 1983 IBR price list and a conversion factor of 1.5 USD per GBP in 1983. The one end of the scale was a 2-user PS10 system with (probably) 512 kB RAM, 10 MB harddisk, a 2-port serial card and the multi-user upgrade for the operating system. This would cost around 13000$, or 6500$ per user (quite a difference to the promised price of 5000$ for the smallest initial model). The other end of the scale would be a 10 user XP20 system with 2 MB of RAM, a 20 MB harddisk, two 4-port serial cards and 9 Fortune terminals. This would cost 38’000$ or 3’800$ per user. We abstract from the fact that 10 users would put quite a strain on the system and that you also would need to buy some more application software, but these prices are probably comparable among different computer systems. If we now compare this price to a quite cheap competitor, the TRS-80 Model 16 with 3 users (1984, 512 kB RAM, 48 MB harddisk) at about 5600$ per user we can see that at least initially, the prices per user were competitive. On a one-user-per-computer-scale, though, the TRS80 Model 16 in 1983 and 1984 would be cheaper before on this market the original Macintosh ($2495), and in 1985 e.g. the Atari ST ($800) would crush the prices and offer a much better user interface. However, Fortune Systems never tried to be in this market.

The CEO leaves the company

At the 1983 new models announcement, Gary Friedman, the CEO also announced that he was to leave the company. Gary Friedman co-founded the company, secured two rounds of venture capital funding. The first round of venture capital in October 1981 added up to 8.5M$, and the second round in May 1982 of 10.5M$. In addition, Thomson-CSF added in May 1982 1.5M$, “reportedly the largest commitment of venture capital ever made to a micro-computer company”. He also brought Fortune Systems on the stock market in March of 1983. There were 5 million shares sold which brought in a sum of 110 million dollars. In 1982 Fortune had a umsatz of about 26 M$ and a profit of 2.9 M$. In October 1983, Gary Friedman resigns on a “disagreement with the board of directors over management style.” Gary Friedman says that “if I didn’t [resign], I probably would have been fired.”. It was rumoured that the companies that put money into Fortune were not happy with the CEO anymore and that the initial technical difficulties of the 32:16 machines (some said that pratictically note more than two terminal could be supported) led to his demise.

SX family

In 1985, Fortune announces the 32:16 SX family. This time the 68000 runs at 12MHz and is announced to support up to 24 terminals. The maximum RAM is still 2MB (so I guess it has still 4 memory slots). A basic SX45 model gives you 1MB of RAM, 45MB harddisk and costs $12995. It seems that the integrated text terminal, the display, and the keyboard are not part of the packet anymore, so you had to buy a terminal with the machine before you could even set it up.

Fortune often sells to bigger companies. In August 1985 InfoWorld reports that nearly 1000 Fortune systems are both installed at Ford and Bell South.

Formula 8000 family


In 1986, Fortune announces the Formula family. They are 68020-based and actually use a new (tower) case! They are claimed of supporting up to 80 users. There are two versions. The lower version for $21900 has 1MB RAM, 70 MB harddisk, a 70 MB streamer. The higher version for $24900 has a 70 MB harddisk. In the announcements, the models are simply called Formula, but later on they are named Formula 8000 series.

Selling the business to SCI

In 1987 Fortune sells its hardware business to the much bigger SCI Systems for between $17M and $20M. SCI has an own line of Unix-based computers, but is more in the scientific and military business. SCI keeps the Fortune brand until about 1988. SCI still exists today.

Formula 4000 family

Also in 1987 a lower-end Formula family is announced. The Formula 4000 family has the same processor at the same speed as the Formula 8000 series, but is aimed at supporting 22 users at maximum. The entry-level 40MB harddisk configuration starts at $9900. The high-end 145MB harddisk, 4MB RAM, 60MB tape drive configuration is at $19900.


So, to conclude the company history, what do we find? We see a company with the right idea at the right time. We see that technical difficulties let the company access the market too late with the competion already at their toes. We see that the company from an investor’s point of view is not up to the promise and burns a lot of money very fast. As a consequence the CEO is fired, and the company falls behind the competition over time, selling the business for a small amount to a competitor.


What I’m seriously confused about are the graphics capabilities of the 32:16 series, or, to be more exact, the lack thereof. Different sources state different things. [Lisa] assumes a resolution of 640 x 480 in “Graphics Mode”, but says that the standard display is capable only of text. [oldcom] speaks about an “optional High resolution graphic card”, but I never saw one appearing in a price list. [CC] does not list any graphics capability or even an option for any Fortune model. Fortune’s own prospectus in 1983, [FS1983a], claims that “Two options may be added to the standard video display controller. First, a bit-mapped graphics display controller with its own 64 Kilobyte memory provides 640 x 480 and 800 x 480 high resolution graphics on the standard 12 inch monochrome display. The second option provides additional memory (256 Kilobytes total) and a second MC68000 microprocessor which is used to drive either the monochrome or color display. This option allows the operator to select 16 colors from a palette of 512 with resolutions up to 1024 x 1024 in the pan mode.“ The very official UK November 1983 price list ([IBR1983]) does not know anything about graphics options at all. So, my guess is that Fortune always promised a graphics option, but never delivered, maybe also because the market never really wanted it. There is no doubt, though, that you could connect (serially) a terminal with some graphics capability from a 3rd party vendor.

From a collector’s point of view

Let’s also talk about Fortune System machines from a collector’s point of view. Text-only 68000-based Unix machines are typically not something collector’s are especially excited about. They are not home computers, they are professional machines, they are not the first in any relevant aspect, and their performance is not extremely good. Probably noone has ever desired to have one in their heydays. On the plus side, the models are all very rare, especially the later ones. They are quite self-contained, so if the main box is running, you probably can use the entire system if you have a monitor and a keyboard or a serial terminal for the later models. It’s a friendly, nice-looking system, and it is one of the earliest 68000 systems you can find. Finally, at that has been always the slogan of this blog anyway, if you have one, even your snobby collector friend will admit that he has never seen or heard about it before. And isn’t that the best reason of all, to have exclusive bragging rights?


As the company existed only for some years, the number of models they produced is limited.


The initial model, presented in 1981, but not delivered before 1983, was called 32:16 after the 68000 CPU which was internally a 32 bit design with a 16 bit bus system. Unfortunately, this was not only the name of the first model, but also the name of the entire model family, so sometimes it is hard decide what is meant when 32:16 is mentioned.

CPU: 68000@6MHz
RAM: 128kB – 2MB (4 memory slots)
ROM: 16 kB
Floppy: 800 kB
Harddisk: 5 – 20 MB
OS: Unix
Text: 80 x 24
Graphics: none
Extension slots: 5
Interfaces: 1 serial port (RS232C), 1 printer port (centronics), 1 keyboard port, 1 display interface
Initial prices: $5000 (for an useless one-user model  without harddisk) to $9990

Thomson Micromega 32

Thomson CSF, the (in 1982) newly nationalized French electronics company participated in financing Fortune Systems from very early on. In 1982, they invested $1.5M and received also the rights to market Fortune’s computer in France exclusively. They did this under their own brand and using their own model name, the Micromega 32 (there has been a Micromega 16 of a completely different design). It also seems to have used a different (probably French) keyboard. A user at ( remembers: “I worked on this computer in the 80s in a french government ministry where it was used for word processing and custom applications based on the Informix database. French ambassies either had Micromegas (for large ones) or Bull PC clones (for the small ones) that would enable them to run the software we wrote based on the Informix database, as this RDBMS was available on Unix (Micromega) and MS-DOS (Bull Micral 30).”

The data of the Micromega 32 (which was sold from 1983) were the same as the 32:16.


These series was designed to be non-upgradable to multi-users, i.e. strictly single-user only. 5 – 20 MB of harddisks were available, the RAM was 256 kB. I never have seen these models mentioned other than in the IBR price list of 1983, so I do not know whether they were really offered.

32:16 PS

CPU: 68000@6MHz
RAM: 384kB – 1MB (2 memory slots)
ROM: 16 kB
Floppy: 800 kB
Harddisk: 10 – 20 MB
OS: Unix
Text: 80 x 24
Graphics: none
Extension slots: 3
Interfaces: 1 serial port (RS232C), 1 printer port (centronics), 1 keyboard port, 1 display interface
Initial prices: $7500 – $10000

32:16 XP

CPU: 68000@6MHz
RAM: 512kB – 2MB (4 memory slots)
ROM: 16 kB
Floppy: 800 kB
Harddisk: 20 – 30 MB
OS: Unix
Text: 80 x 24
Graphics: none
Extension slots: 5
Interfaces: 1 serial port (RS232C), 1 printer port (centronics), 1 keyboard port, 1 display interface
Initial prices: $17000 – $19500

32:16 SX

CPU: 68000@12MHz
RAM: 1MB – 2MB (4 memory slots)
Floppy: 800 kB
Harddisk: 45 – 70 MB
OS: For:Pro
Graphics: none
Extension slots: 5
Interfaces: : 1 serial port (RS232C), 1 printer port (centronics)
Initial prices: $12995 – $14995

Formula 8000

CPU: 68020@16MHz
RAM: 1MB – 4MB
Floppy: 800 kB
Harddisk: 70 – 145 MB
OS: For:Pro 3.0 or Unix System V.2
Graphics: ??
Extension slots: 5
Interfaces: : 4 serial ports, SCSI
Initial prices: $21900 – $24900

Formula 4000

This is the first model after the hardware business is sold to SCI.

CPU: 68020@16MHz
RAM: 1MB – 4MB
Floppy: 800 kB
Harddisk: 40 – 145 MB
OS: For:Pro 3.0 or Unix System V.2
Graphics: ??
Extension slots: 3
Interfaces: : 4 serial ports, SCSI
Initial prices: $9900 – $19900

Later models

Fortune/SCI had a last series of models in 1990. The FORTUNE/SCI System 5000 family had an Intel 80×86-based low end (using Unix) and a high end that added a Motorola 88000 CPU to the system. But in my opinion, that were SCI, not Fortune models, therefore I omit them here.


[CC] CC-Computerarchiv,
[Lisa] Lisa Sales Marketing Binder, June 1983
[IBR1983] Fortune 32:16 Microcomputer Systems Suggested Retail Price List effective 1 November, 1983. IBR Microcomputers Limited.
[FS1983a] Fortune Systems. Fortune 32:16. The Complete Business System. 1983
[Warnock 2004] comp.unix.bsd.freebsd.misc, 26 May 2004, Rob Warnock, Re: Andrew Tanenbaum on the origins of Unix/Linux
[ClassicCMP2005] 2005 posting to ClassicCmp cited after

34 Responses to “Fortune Systems”

  1. joebarflyJoe Clarizio Says:

    Wow, this brings back memories. I was an authorized Fortune technician working for Computer Land in 1982/83. I was also an authorized Apple Lisa tech. Didn’t really do much work on either.

  2. David Watson Says:

    I implemented a complete accounting system and marketing system using the Sculptor 4GL, linking the main Fortune 32:16 to about 8 Wyse terminals … back in about 1984/1985 … don’t remember the exact dates.

    Loved the machines … it worked really well.

  3. Leonard Bernal Says:

    Wow. Old memories. I worked there from 1981 to 1983 as a draftsman. I did all the technical drawings and some minor designs for the 32:16 enclosure. I also got my start as a PCB designer doing a Voltage Sense Lab board design. I still have the prototype in my box of boards up in the attic. -Leonard.

  4. John L. Bass Says:

    I joined Fortune Systems in Jan or Feb 1981 as Employee #5, after the three founders and Rob Warnock, working out of Gary’s office in the Embarcadero Center down-town San Francisco.

    Rob and I worked out the basic architecture as a low profile desktop based on the new Motorola 68000, integral MFM hard drive/floppy, running UNIX and business software based on the Wang and Microsoft like offerings. I took my Radio Shack TRS-1 up to the office so we had a spreadsheet and word processor for the early business plan and budgeting cycles.

    As soon as we had an initial investment commitment, we picked up a warehouse office a few blocks away off Broadway, where we assembled the barebone initial sales, marketing, and engineering teams.

    There was a hard $9,999.00 cap on the initial baseline system for marketing reasons. Unix V7, 256K memory, 10MB hard drive, floppy, and word processor — and margins for two-tier distribution. We had a rough motherboard design, enough to start freezing mechanicals.

    A plexiglass mock up of the case, keyboard, and monitor was done, that became the signature 32:16 series that we used for dog and pony show to close initial investors.

    A few months later we picked up our first real building in Foster City to house a staff of about 35, and initial manufacturing space. Prototypes were knocked out for development, both in house, and for our major investors/customers Thompson CSF, Southern Bell, and a few others. We knocked out core software offerings, UNIX, Wang like word processor, Spreadsheet, etc … enough to close Computer Land, and some other large retail outlets and vars.

    There was a mythical man month blunder by Homer Dunn and Rick Kiessig that cost the company about 18 months in schedule before the UNIX V7 port was stable enough for retail end customers. This paralleled a cost reduction mistake from the initial Western Electric power supplies to Zenith power supplies, that introduced severe magnetic flux noise into the hard drive causing random system failures that were difficult to narrow down with an already unstable operating system. Rick’s blunders cost the company severely, and pushed the companies profitability back nearly two years, opening up competitive pressures that should never have been an issue.

    The company IPO was a market record of a little over $100M (some sources say $120M) that wasn’t broken again for years. Shortly after that, the US and world economies dived into the 1983 recession. While many of Fortune System’s competitors were quickly insolvent, and in bankruptcy, Fortune’s sales just went flat (dropping from a crazy exponential market introduction curve). That was enough for the new Wall Street lead board, to quickly purge the founding executive team for failing to meet market/sales projections.

    The board replaced the executive team with an old school big iron group (I remember Boroughs, but it was a long time ago). With nearly all debts paid, a lot of money in the bank from the IPO, the new leadership halted all new product development, and cut staffing back to bare bones to curb some very minor losses. When the market recovered 2 years later, Fortune had a 5 year old design that was no longer competitive against a huge number of new competitors that spent the market down turn design new state of the art products. This single stupid decision cost Fortune it’s entire market lead, and ultimate killed the company.

    The Formula product line was two years late, and 3 times the market cost expectations as Sun and other Desktop UNIX companies took over the market.

    This was due to the grossly incompetent second round management team that was still thinking in big iron markets.

    Many of us early engineers left when the no new product development mandate happened, and projects were cancelled. Some of us independently completed major 32:16 projects externally, and sold the designs back to the company because the in house engineering teams lacked the support and experience to complete 32:16 projects that were committed to for major customers.

    I did a SCSI drop in replacement for the MFM controller. Another engineer did a bit mapped graphics card. Another engineer did a smart serial card. All these projects pissed off the engineers in house, each wanting to start a NEW project …. but the no new product development mandate continued, with the exception of the early Formula M68020 development work.

    Marketing saw the high end (big iron) Formula product as the wrong product at the wrong price. At their request I had my SCSI board team design a bare bones M68020 desktop system, shoebox sized with SCSI HD, Floppy, 1-4MB of memory, and 24 serial ports on a smart serial controller … meeting their tight $1K hardware cost budget, for a $3k market price target. Took my team 3 months to turn this prototype, including a UNIX port and drivers. Marketing was blown away when we took the prototype up for demo, meeting their price target, and being 20% faster than the Formula prototypes. Engineering went off the rockers crazy when marketing pushed to bring another outside project in house. Engineering gave a short list of “problems” that didn’t meet their “standards”, and firmly said NO WAY IN HELL. If markeiting wanted a low cost project, they would design one, AFTER the Formula release … about a year out.

    At marketings request I took the fix it list back to my team, and we rolled a completely new design that addressed all engineering’s requirements, and then some … including a few more wish list items from marketing.

    Three months later the round two prototypes had a running UNIX port, with a completely redesigned memory and IO system, that far exceeded requirements. 4-16MB memory system, smart SCSI IO for HD, tape, and foppy, plus smart IO for serial ports … M68020, with floating point processor, and two independent IO co-processors sharing memory. And still 20% faster than the Formula.

    Marketing was blown away with our short turnaround complete redesign and pushed hard for engineering acceptance … after some major in-fighting market got turned down as engineering threatened to walk out. Our demo exceeded marketings requirements, but we walked away without a contract or sale for trhe project … and got a legal notice to cease claiming full rights to the design by Fortune, so we couldn’t even take it to another buyer. A lot of money down the drain, following marketings lead on that project.

    The in fighting we caused with the shoebox Formula project caused the company to explode in the following year. Marketing sold the 32:16 and Formula hardware product lines and engineering to SCI since they couldn’t control engineering to meet market needs.

    Suzy took the inhouse software teams with the business software products, and spun off as a software company.

    Marketing settled outstanding stock holder lawsuits, and used the remaining cash to explore venture capital startups.

    Others have different views on some of this … but this is mine.

  5. John L. Bass Says:

    Comment on Mysteries … At the point in time the 32:16 was designed in 1981, the available CPU’s and memory subsystems were in the 1 mips range, not nearly enough cycles to handle graphics with a significant resolution and color depth. And certainly not multi-user while doing so. The $10k+ price was half the starting annual salary of an engineering graduate, so ASCII terminals extended the seats per system into the 4-16 range to get the cost per user down below the price of an IBM selectric typewriter which cost about $2k/yr including lease/maintenance contract at the time.

    Word processing seats sold dominated other uses by nearly two orders of magnitude, strictly because of the high cost of ownership/use of a Selectric Typewriter. Point of sale, with ASCII terminals, was a distant second use.

    The Lisa was too expensive and slow at the time … something of a low production research machine, as the only other production bitmap machine was the Xerox Alto at the time … another expensive ($32k+) low production research machine.

    It wasn’t until 1983 when memory prices started dropping during the recession (inventory glut) that bit mapped machines became affordable, and we saw the first affordable ($2.5k) MacIntoshes hit the market in early 1984. And things really took off with the much faster M68020 products in 1985.

    This is the time period that Fortune got it’s graphics card from either Hub or Rob, but there wasn’t really software to drive it as the main core software was ALL ASCII terminal based, with block graphics.

    Note that 10mbps 10bT Ethernet was still 6 years away from becoming commodity at this point, and most people still thought of Ethernet as 3mbps coax with Tat Lam taps. I think it was Rob that did an Ethernet card after leaving Fortune, and Hub did the bitmap card after he left

    Everyone was still using fast RS232 for networking links between machines, including UUCP and early SLIP during the 1980’s

  6. John L. Bass Says:

    MMU designs and software security pal comments

    The MMU style used by most machines prior to 1985 was segmented memory, made really popular by the Dec PDP11 series machines, on which UNIX was developed.

    Demand paging that we see in microprocessors after 1985, was rare before then, and only on very large machines for the day. The Dec Vax series was the first machine in the $250k range that supported demand paging, where PDP11’s running UNIX with segmented MMU’swere in the $100-250K range.

    The first microprocessor UNIX system we did at Onyx Systems in 1980 also used a custom segmented MMU, the mimiced the PDP11 design to make porting UNIX V7 easy.

    And for the same UNIX V7 porting reasons, I drove the Fortune System 32:16 architecture to also use a segmented MMU. We had a very tight schedule to deliver working prototypes by the Fall COMDEX show in 1981. Anybody that had a working PDP11 UNIX application needed to be able to port it to the 32:16 with minimum trouble (other than the bit order issues between the PDP11 and M68K).

    Segmented MMU’s have a significant performance advantage when using DRAM, in that you can overlap the segment lookup (using high speed static RAM’s) with the DRAM RAS cycle, and then present the table lookup result as CAS.

    When I did the shoebox Formula design, it was also segmented for these reasons …. easily port PDP11 and 32:16 OS/applications, plus had a 1-2 cycle speed advantage over M68020 designs using the Morotola M68851 Paging MMU.

    Larger applications require a paging MMU to gain a larger usable address space.

    The Fortune Systems Anti-Piracy system was driven by Software VAR’s that had been struggling with software theft in the Z80 and Apple markets. There is a PAL on the motherboard that generates several puesdo random numbers that are the machines software key. When a virgin software disk is installed on a machine for the first time, it becomes branded to the machine, and can not be re-installed on another machine.

    As a skilled hacker I was very skeptical of the implementation, and specifically refused to sign the internal NDA regarding it’s implementation which was delivered by an outside security contractor.

    It’s my understanding that there are only about 2^10 unique bits in the machine ID, so there are really only about 1024 unique machine ID’s even though the number of bits in the puesdo random number are more. The theory was that even a large customer would be unlikely to have very many out of multiple machines with the same effective machine ID.

    I’ve always guessed that the security of the installation media required hiding some bits in the floppy sector headers, which wouldn’t take much to reverse engineer, but be just beyond what a normal IT person could do.

    When I designed the shoebox Formula system I knew that I would need a security pal on it for marketing, so I contracted with a friend (Diller Ryan) to give me a clean room design specification for the security pal socket. I gave him a 32:16 and an HP Logic Analyser, and he came back in a couple days with the clean room specification for the socket, and how to read the security PAL.

    For testing purposes I had previously used a logic analyzer to locate the Fortune System kernel’s routine that read and validated the security pal, and returned the machine ID. I patched a kernel to always return the same machine ID, that didn’t need a security PAL in the socket. This is all that is necessary to defeat their security scheme. Takes about an hour for a relatively skilled systems level kernel engineer., Once you do this, a copy of that disk image will run on any machine.

    As a side note, I believe that machine ID zero, is universal and will run any software without checking for it’s ID. This is what was used internally for QA and engineering …. and is the mastered ID for un-installed software. Just a guess … but I’m pretty sure.

    there are also a couple other ID’s returned by the PAL … I believe one of those is the vendor ID for the software creator, and another is for a facility ID, where for certain large customers any machine with the same facility ID can run any copy of the software. I didn’t dig very deeply into how the PAL results were used, so these are just guesses.

  7. John L. Bass Says:

    If you have a Fortune 32:16 with a working kernel, the base software is UNIX v7 with only a few significant changes. The kernel is a BSD2.x variant, that uses the Berkeley filesystem. The Fortune C compiler is an MIT M68K PCC variant.

    The floppy chip has a straight forward interface, and is CPU driven. DItto for the serial and parallel interfaces. Hook a logic analyzer to the CPU and various interfaces, and you will see how they are decoded.

    The MFM WD controller is a variant of the Western Digital 1000 series chip sets … you can reverse engineer the driver from the roms on the card … pretty small.

    When I did the demo OS for the shoebox Formula system, I took stock UNIX V7 with some BSD2.x patches, using the MIT M68K PCC complier tools. Took about 6 weeks to port and get it stable.

    Will take a small bit longer to reverse engineer a working system, and port stock UNIX v7 using the MIT M68k PCC compiler.

    Sources for UNIX V7 and BSD2.x are in the ancient UNIX archives, and are open/free to use. May take some digging to locate the MIT M68K sources.

    All this stuff is TINY and extremely simple by today’s software engineering standards. EVERY program compiled, including statically linked libraries is less than 64KBytes … and running systems were 128KBytes ram with 5MByte disk drives. NOT GBytes … just KBytes and MBytes. Small enough that if you have questions, and can not find sources, just toss a disassembler at it, and reverse engineer with ease.

  8. John L. Bass Says:

    FYI … when Fortune Systems closed one of their facilities after SCI bought them, I purchased at the auction a pallet with a couple dozen boxes of 9trk backup tapes from the PDP11 internal development system (I used to manage) that was used during 1981-1983. I’m pretty sure they were not erased, and should contain archives of internal email during that period, along with sccs archives of the development sources. There will probably be errors, maybe lots of them, but with multiple full backup copies over the period, should be able to reconstruct nearly everything for that period, prior to the IPO.

    Post IPO, with local tape drives, most development archives became more local, and didn’t archive on the PDP11 machine.

    The tapes have been sitting in a cool dark closet in my basement, waiting to find a really good 1600bpi tape drive to recover them. But hard to tell what 36 years of bit rot may have done to the tapes.

    I picked up a later 32:16 on ebay last month, with 512KB ram, and a dead hard drive … less keyboard cable. Got two FS terminals too., less keyboards. System seems fine otherwise, as it progresses to E05 on the basic start up diagnostics. SO at least the CPU, memory, and video are functional. Was happy to see that it had a WECO power supply, instead of the Zenith. Need to find two keyboards, and 3 keyboard cables.

    Once I can find a keyboard cable, I can probably start working on finding a way to get it to boot something.

    Would be nice to find an ex-engineer that kept a set of schematics for the machine … I was too careful to purge everything as I left Fortune.

  9. John L. Bass Says:

    FYI …

  10. John L. Bass Says:

    Comments about staffing (and project management) done right/wrong at Fortune.

    The three primary founders, Gary Friedman, Dave Vanderberg, and Homer Dunn (sorry if I miss spelled any of those) created a policy from the start, of hiring from the top, down the org chart.

    Specifically requiring that you do the job of your team, until funding was available to hire your team. This pretty much held for the first two years. People that got hired were actually capable, rather than having a strong Peter Principle resume. It also stopped dead in it’s tracks, fighting over promotions as new positions opened up … new positions were always down the org chart.

    I’ve worked in quite a few positions, in several dozen companies over my 50 year career. The two most outstanding teams were SRI International and Fortune Systems. Both with exceptional staff, top of the line.

    The critical exception to this was my hiring, and a stand-off 5 months from start. They needed a top UNIX Systems Programmer lead, and I was available after just completing the UNIX V7 port to the Zilog Z8000 microprocessor at Onyx Systems, with a crushing 3 month schedule, The number of people with cross architecture UNIX porting experience at that time could all be counted with the fingers on one hand, and there would still be a finger or two free. Homer and Rob caught me after giving a talk at the UNIOPS conference, gave me a great offer, and I accepted. Signed the papers the next day, or so.

    A couple weeks later they hired my boss, the VP of Engineering (whose name I can’t remember right now). We both hit it off well, sharing a strong KISS design methodology, with regular incremental deliverables and early rigorous testing. I was told I would have funding for a staff of 7-10 in a few weeks, so I spent the next month lining up the best UNIX systems programmers with strong experience. My teams hiring got pushed off week by week for four months, to the point that several people on my hire list took other offers.

    And to make things worse, marketing was demanding additional deliverables that were high risk and poorly defined. Tight schedule KISS is one beast, open ended research projects with a deadline is hell.

    My boss and I were at wits end, having a hard deliverable of fall COMDEX rapidly approaching, and becoming way over committed schedule wise. We drew a line in the sand with Homer, and said the schedule would slip day by day until I had staff hired. We were both replaced a few days later.

    Homer kept a plaque on his desk that said something like, The man that says NO, is the WRONG man for the job. We said NO, so we got replaced with YES men/boys. Nobody else would commit to delivering the UNIX port in 5 months, especially when my team wouldn’t step up without me, and they were the best.

    So the new VP of Engineering showed his colors in the first week, giving the investors a dog and pony show, and openly lying about the state of our progress … we were told to shut our mouths and not say anything. That day I was introduced to my two replacements, Huba Toth (a quite capable project manager, with a strong QA background) and Rick Kiesigg who’s resume claimed he had completed three cross architecture Motorola M68K ports. And the VP introduced Rick to the investors with those credentials in front of the engineering team.

    I was well plugged into the UNIX world, and I knew there wasn’t any project farther along than Fortune’s. Not even the stealth projects.

    I was pissed, but kept my mouth shut, still trying to sort out my standing, and possible career changes. I was introduced to several college kids late that afternoon that were joining the project. Left early, went to dinner with my fiancée, and tried to cool down.

    Went to work the next day, and went back to working with a couple hardware engineers bringing up a new prototype board. Rick found me not long after, and started asking me what he needed to do. What did the team need to be doing. Pissed me off after taking my job, so I told him he was the expert, since he had done three cross architecture ports and I had only done one. In front of the others I point blank asked him what they had done on the other three ports. His answer was, they never got that far, they were just paper exercises waiting for the real hardware prototypes to be built.

    I went to Huba, unloaded, and left for the day. I requested a new job assignment, directly under Huba, and out from under Rick. I had that when I came back in the next morning. Shortly after that I took a position in mfg engineering to deliver mfg and field service diagnostics … tools for the line techs to checkout systems, and tools for the field to diagnose failures. On my new office wall I kept a pert chart for the OS completion. I extended that chart with every new feature Marketing committed them to. Started out a 12 months over schedule, and ended up 24 months over schedule … I was right within a few weeks.

    If you haven’t figured it out already …. NEVER pad your resume, as someone might actually expect that you have the experience to do the job you were hired to do.

    Rick hired nearly two dozen college kids, some recent grads, some still in school. The UNIX port became a grand research project, with every new wiz bang feature you could imagine. If Bass could pull of the Onyx Z8000 port in 3 months of long hours, then it would be a piece of cake for 20+ kids to finish the Fortune port in 5 months … no problem.

    Welcome to the reality of the Mythical Man month. NOTHING was stable the weekend before COMDEX, so the kids all went to Tahoe skiing. Huba and I, plus a couple other decent engineers spent the weekend hand crafting a demo for marketing to give at COMDEX …. with strict orders NOT to deviate from the demo script, and NEVER let a customer touch the demo machine.

    COMDEX was a roaring success, Fortune was the best of the show. But Fortune didn’t take a product to COMDEX, it took an automated storyboard that required careful scripting to work

    The kids took a lot of credit, thinking they were almost done, completely unaware of the 90/10 rule …. 90% of the features, take the first 10% of the schedule …. the last 10% of the critical features will take 90% of the schedule.

    Three months later, and nothing was stable, and the kids blew through the first critical customer deliverables to major investors. They made a lot of promises, took on new marketing wish list commitments in exchange to schedule slip … and several months later blew through that deadline as well.

    At a year past schedule, and facing another COMDEX, Homer called me into his office with Huba, and asked me to take control of the systems team, replacing Rick. Homer wanted to fire the entire team, and bring in a fresh experienced team. I declined, and defended the team of kids, telling Homer he had just spent over $3B in lost sales to build one of the most experience UNIX teams in the valley, and that it would be stupid as hell to fire them. The revenue losses of 150,000 machines from the first three years product introduction curve, with average sales of $27k/unit with terminals and software is NOT a small number. They took my recommendation, and hired an excellent people oriented project manager, fired Rick, and recovered the project with some delicate care handling the kids that were completely demoralized by that time.

    Thus enters yet another validation of the Mythical Man Month. 20 really good kids with open ended controls, will never complete what 7 excellent experienced developers will do in the same time frame. 7×5 is 35 man months. 20×18 is 360 man months.

    They paid Rick significantly more than me to say YES, but that was insignificant to the organization cost of the lost sales.

    We recovered … but it was a $3B loss at best, and in reality lead to the leadership changes that buried the company in the end.

    Just say NO … do not compromise profession integrity for short term gains.

  11. John L. Bass Says:

    Quality, Quality, Quality, in the face of High Risk, and more Risk.

    Ignoring Homer’s and Rick’s mythical man month blunder, I started out saying that Fortune Systems had one of the highest quality and experienced staff of any company I’ve ever worked for.

    The founders built from the ground up, a computer company ready to take on IBM in the market. The highest quality standards in sales, marketing, training, sales support, HR, engineering, tech pubs, manufacturing, and field support staff and engineers. NONE of this full established company staffing was an after thought, as every group was staffed from the start, top down, with highly experienced people. This was a mind blowing change from other bare bones bootstrapped start-ups I had worked with, like Onyx Systems where all the back end functions never materialized.

    The documents roadmap appeared before we even had the product defined. And I mean marketing documents, sales literature/document, national and international advertising, training documents, VAR documents, Reseller documents, User documents, all the way out to field service documents. All of these documents took on a life before the product existed, leading the product definitions and specs, and tracking development right up to QA release.I had been working the the computer field for 14 years at that point, and the ONLY company that was that comprehensive with literature and manuals was IBM.

    So I would like to make the point again, the founders built this company from the ground up to take on IBM at it’s best game, up and fully running from day one prior to product release. Never was trying to compete on par an after thought … it was designed into the company from day one.

    The training staff was out in the field training Computer Land (and other retail outlets and vars) owners and staff how to use, sell, train, and support customers, long before the machines were ready for volume ship. Fortune never considered letting the market drive sales and acceptance … Fortune was in front of the market, and driving it from day one.

    Every other silicon valley computer company did it backwards, and the state of the computer market reflected it. If you haven’t done so, watch the movies about Apple as a start up, boot strapping with nothing but an idea. That was how Silicon Valley computer start ups did business … and Fortune was way ahead of the game.

    Every other computer start up also boot strapped manufacturing, with the life and death for revenue leading decisions to avoid Chapter 11 (or 7). Expensive things like Quality Controls came last.

    Manufacturing at Fortune started with strict Quality controls, well before the first systems were built. Full component evaluation standards up front prior to vendor acceptance, QC acceptance checking for incoming parts, QC down the line, all the way out the door. This is how Western Electric does business, not Silicon Valley start ups.

    Every disk drive we qualified went through full environmental testing, in many cases exposing critical defects in design the vendor was expecting to slide past our engineers. Subtle things like write hot and read cold failures, or write cold and read hot failures, self induced vibration failures, shock/vibration isolation failures from multiple drives on the same chassis.

    It was not an accident that the used machine I bought off ebay last month actually was functional when I powered it up 35 years after it left the manufacturing line. This was the direct result of the rigorous Quality controls in the product all the way to ship. Yes the disk drive is dead … everything else was functional.

    In comparison, every China manufactured product I have worked with in the last 20 years, has an infant mortality rate of between 5-10% in the first 6 months (leading edge of bath tub curve). Most of these products also has the back end of the bath tub curve cutting in between 6-12 years, with 80% of the product failing inside a few years after that. Designed for obsolescence.

    I acquired sixty (60) Rackable Systems servers about 15 years ago for a computational cloud server. About half them failed by year 8, and today only one remains functional. Power supplies dead. Motherboards dead. I’ve mixed and matched right up to the last one, including buying a number of power supplies to get this far.

    At Fortune, the VP of Manufacturing brought in several quality experts during the initial prototype design process, and instilled QC as a design discipline from day one. It became a fundamental standard inside engineering.

    Western Digital protested loudly when our engineers documented clear design failures in the WD1000 chipset data separator, that failed to meet even their own jitter specification. Ditto for memory vendors that had serious Alpha particle problems in the chip plastic that were caught by Fortune QC. Nearly every hard drive vendor complained bitterly that we screened drives to their own specifications, and failed significant quantities of drives. Other components failed accelerated life testing, and were disqualified, and in some cases we completely disqualified every product from that vendor because their products did not even meet their stated final QC acceptance prior to ship.

    Why was this important?

    It was pretty simple … nearly every component in the Fortune System design was bleading edge technology, and had not yet reached it’s own natural maturity cycle.

    That was a HUGE risk for Fortune Systems, and the founders built a company to integrate that risk, and mitigate the risk to near zero with industry leading QC systems from front to back.

    Nearly all the logic on Fortune Systems boards is programmable logic — AKA PALs. Fortune was one of the first companies to take the leap of faith (backed up with rigorous QC) and build a product line around this technology. And we got bit in the process, but QC screened those problems before they were shipped, and we worked with the vendors to fix them.

    I took this even farther than Fortune with the shoebox Formula design, shrinking the design down even more so that it really did fit in a shoebox.

    Quality and Risk mitigation set Fortune Systems apart from every other Silicon Valley computer company.

    That Fortune Systems failed, due to two simple human factor mistakes, is still mind boggling to me today.

    • Mike Friese Says:

      I worked for Western Digital. I designed the parallel section and firmware for the hard disk controller on the Fortune 32:16. The Mr. Bass’ comments about the jitter in the data separator were a refection of a controversy going on inside WD. Engineering had designed a phase detector (pump) using STTL that performed well but we were pressured to integrate that into the 1100 chipset. The logic of the NMOS gate arrays was just not fast enough to do the job properly with our particular design.

  12. John L. Bass Says:

    When we took the 32:16 to Fall Comdex in 1981, the only other M68K machine that was near market release was the Tandy TRS Model 16, which also started shipping in low volume in the spring of 1982.

    By summer of 1982 Tandy had shipped about 2,000 of the machines to stores, with only a few sales because there wasn’t any application software to run on it’s 16bit version of TRSDOS.

    Fortune had shipped more that with beta versions of UNIX and beta versions of all the core applications, with restricted distribution to core customers, resellers and VARs while it struggled to resolve the OS stability issues and hard drive errors.

    The NuMetal shields solved a lot of the hard drive problems with the Zenith power supplies, and a data separator redesign solved the Western Digital WD1000 chipset problems. We did a piggy back data separator to fix the existing WD1000 MFM controller cards, and pushed WD to fix the chipset, and rolled a new version of the controller a few months later that was stable.

    By the fall, with the hard drive problems under control, the system was stable enough that with a single user doing normal word processing or development work, it would only crash a few times a month. With heavy load/stress testing in the lab, it generally took less than a day to crash the machine, if not a few hours. Still only beta quality, but very usable. Homer made the leadership changes to bring the mythical man month blunder under control.

    Sun Microsystems started shipping the Sun-1 hardware with a basic UNIX port.

    Apple introduced the Lisa in Jan 1983 at the same $10k price range, and it’s own set of significant problems and challenges. The profile hard drive was crazy slow, connected over a slow speed interface. There wasn’t any significant software available.

    By spring 1983 the 32:16 OS was a lot more stable, and the Fortune development teams were able to complete and clear a back log of critical but lower priority errors, and committed features that had been delayed. The backlog of orders started turning into real shipment volumes. This allowed Fortune to proceed with it’s IPO plans in March.

    And then the 1981-1982 recession finally hit technologies and Fortune Systems, Fortune sales went flat, still shipping but well off forecast.

    Fortune took a huge hit in the press as tech stocks crashed in the market during 1983, including the new Fortune stock.

    One major paper headlined Osborne, Victor, Fortune Systems file bankruptcy … the first two did, but Fortune never did, not even close. There was a lot of vile words about Fortune, doing a record IPO right as the recession hit technology stocks.

    And that certainly lead to Gary’s replacement, along with a number of other senior exec’s and managers, and a new managemernt team came on board.

    I was told that Gary was the fall guy at Itel, and had never been responsible for it’s fail. Gary had been a founder of the company, but had been pushed out when he refused to accelerate growth with heavy financing. When the market collapsed, and the financing the exec’s used after he left took the company insolvent, he stepped back in to try and rescue the company with restructuring and refinancing, but failed to assemble investors to take the risk. And in the end became the fall guy, since he was at the helm in the end.

    Whether that was true or not, I’m not sure. But the heat Gary and Fortune Systems took in the financial press sure had a lot of hate and anger that went well beyond the success at Fortune Systems. I still believe today that the company would have survived and thrived, if Gary had remained for another 2-3 years … and exceeded his visions for the company.

    Gary would have never made the “no new product development” mistake that killed the company … as he had already been pushing for the follow on product lines to the initial 32:16. The company had the money, the talent, and customer base to do it, and do it well.

  13. Wouter Says:

    Wow. Great info Fritz and John. John, if all else fails get those tapes to Jason Scott. I for one would love to see what’s on there.

  14. Barbara Says:

    Wow John, your comments were a super interesting read! I was also at Fortune in those early days, but mostly out in the field as an application systems engineer. I had a completely different perspective on what went on there, so I really enjoyed hearing your view!

  15. John Bass Says:

    After some searching, I found Rob’s replies in the BSD threads:!topic/comp.unix.bsd.freebsd.misc/QGdMwZJOoKY

    Rob Warnock
    david parsons wrote:
    | John wrote:
    | >A memory management unit (MMU) is required to support a
    | >real Unix because of the semantics of fork().
    | Well, not exactly; there are software workarounds for not having
    | a MMU. I think that Minix/68k, for example, got around the lack
    | of a MMU by physically swapping the forked processes in place.

    The Fortune Systems “32:16” was built in 1981 and ran Unix on a plain
    MC68000, *not* a 68010 or 68020 [because they weren’t available yet!!].
    Following the lead of TOPS-10 on the DEC KA10, we used a simple base/limit
    style of MMU, built out of MSI TTL chips (using, IIRC, 74S670 4×4 register
    file chips for the base & limit regs). We chopped up the available 16 MB
    of address space (24 bits was all the original 68000 supported) into four
    segments as follows:

    Segment Base addr Direction of growth
    text 0x000000 upwards
    data 0x400000 upwards
    extra 0xbfffff downwards
    stack 0xffffff downwards

    Only text, data, and stack were really needed for Unix, but since
    hardware register files tend to come with powers-of-two elements, we
    went ahead and implemented the “extra” segment anyway. [For hardware
    simplicity, it grew downwards like the stack seg did.] As expected,
    we did discover some uses for the “extra” seg, such as implementing
    shared-memory data segs or for allowing privileged processes to mmap()
    I/O space [for user-mode hardware drivers].

    Obviously, with such an MMU we couldn’t support “paging”, only
    “swapping”, but that was o.k., since the original 68000 couldn’t
    always recover completely[1] from segfaults anyway! That had to
    wait on the 68010…


    [1] It *could* reliably recover from segfaults that occurred from a
    subset of the instructions, so we were able to safely use segfaults
    to auto-grow the stack by designing the compiler’s subroutine entry
    prolog so that immediately on entry it used one of the “segfault-safe”
    instructions to touch the lowest stack address that the subroutine
    would ever use. This ensured that any (normal) stack faults would
    occur on a recoverable instruction. Worked like a charm, actually!

    Rob Warnock
    627 26th Avenue
    San Mateo, CA 94403 (650)572-2607

    Rob Warnock
    Greg Hennessy wrote:

    | Steve O’Hara-Smith wrote:
    | > Perhaps my memory is failing but ISTR an external MMU for the 68000
    | > being available.
    | I could be wrong, but something tells me that the motorola version
    | had to be paired with an 68010 rather than a plain 68k.

    The *Motorola* MMU chip needed a 68010, but in a parallel article in this
    thread I discuss the TTL/MSI-based Fortune Systems 32:16’s base/limit MMU,
    which supported Unix on a plain 68000…


    Rob Warnock
    627 26th Avenue
    San Mateo, CA 94403 (650)572-2607

    Rob Warnock
    Richard Tobin wrote:
    | Steinar Haug wrote:
    | >Anyway, the 68000 had other problems too – like no instruction restart.

    | I believe some other 68000-based machines attempted to solve the
    | problem by having the compiler never generate code that could cause a
    | page fault except in a simple load or store instruction.

    In a parallel article in this thread I describe exactly that strategy,
    which was used quite successfully to run Unix on the Fortune Systems
    32:16 box (which has a simple multi-segment base/limit MMU). It didn’t
    permit demand paging (or in fact, *any* paging, only swapping), but it
    *did* permit automatic stack growth by controlled segfaults during
    subroutine entry…


    Rob Warnock
    627 26th Avenue
    San Mateo, CA 94403 (650)572-2607

    Rob Warnock
    John wrote:

    | Yes the 68851 PMMU. Also 68881 for math CoPro. Although I believe you
    | needed to pair it with a 68010 which included restartable instructions…

    | No consumer level machines came out with a 68851, that was
    | reserved for workstation class machines like Sun. The first
    | time any Mac, Amiga or Atari had a PMMU was when they started
    | shipping with 68020.

    Though the Fortune Systems 32:16 (see parallel articles in this thread)
    was in some sense a “consumer level” machine [at least for business
    customers], since it was a desktop Unix box with an entry price of $5000.
    It used a proprietary, simple, multi-segment base/limit MMU (“swapping”
    only, no “paging”) with a plain 68000 [since neither the 68010 nor 68020
    were available in 1981 when it was built].


    Rob Warnock
    627 26th Avenue
    San Mateo, CA 94403 (650)572-2607

    Rob Warnock
    Christopher Browne wrote:
    | The 68000 couldn’t do virtual memory; that waited for the 68010.

    That depends on what you mean by “virtual”. See my parallel articles
    in this thread describing the Fortune Systems 32:16 box which used
    a simple base/limit MMU to run Unix on a plain 68000. (The MMU enabled
    “swapping” and “shuffling” and read-only shared-text, but not “paging”).


    Rob Warnock
    627 26th Avenue
    San Mateo, CA 94403 (650)572-2607

  16. John Bass Says:

    In my initial post, I stated “A few months later we picked up our first real building in Foster City”.

    That was wrong. After the San Francisco warehouse, we moved to an industrial building in San Carlos near the Circle Theater, just off Hwy 101. About a year later we moved to the Foster City building.

  17. Ronnie G Says:

    Anyone able to ‘set me straight’. My last few years before retirement I did end user support for Fortune computers deployed in the local phone company in South Floridia. At the time I was told (by someone) that BellSouth bought out all the Fortune inventory. Now I don’t know if that was all the inventory from a local seller or all the inventory from the MFG. Anyone have any knowledge or thoughts on that?
    I worked in the central office environment. The cup’s we got were hand me downs from other departments. Had we not taken them they were headed to recycling.
    I was self trained starting out with Coherent then Linux & Fortune.
    Thanks for any thoughts.
    I really enjoyed reading this post. Not many people know about the Fortune computers. I only stumbled onto this article while looking for a way to add a fortune onto my .sig on Linux

  18. Michael Westmore Says:

    WOW! Thanks for this history lesson. Homer was my uncle. I can remember him taking me to a sales meeting in Sherman Oaks, CA to show me what the company had been working on. He had promised me a 32:16 but now reading about it I guess I see why I never got it.
    John, if you have any more info I’d love to hear it.

  19. Gary Mahalak Says:

    My first Computer Job back in 1983 was installing and fixing Fortune systems running SCO Xenix or SCO Unix. Still have the handbook. We were around Detroit area and eventually Fords took interest but the power supplies were bad and I was always working on them.

  20. Guy Mark Tibbert Says:

    Wow, this brings back memories. I was 20 in 1985 – and we had a Fortune 32:16 (not sure which model now) at Tetra Business Systems in High Wycombe. (Along with Altos 586,986,2086,3068, Motorola Miniframe and a few others).

    The Fortune 32:16 was the first multi-user computer I ever used Until then I had only ever used micros – never “minis”.

    We used to write most of the new code on the Altos machines and the squirt via RS232 if the changes were significant or just edit manually if it was just a minor bug. I for some reason usually squirted to the 32:16 and the Miniframe.

    You are right about the 32:16 keyboard being huge – it was a little quirky in a few ways. I also recall that we had to do something else other than sticking a GETTY on the TTY ports to enable another user – but to say memory of that is vague is a bit of an understatement.

    The company sold Altos kit though – so most of the time, we used those – but the real danger of using the Altos 586 or 986 at the console was the dreaded “tit switch” – the RESET button mounted next to the 5¼ drive – so called because you felt a bit of a tit if you went to eject the disk and pressed the reset button – killing everyone’s processes. In the end, we taped a small plastic pot over the switch so it couldn’t be pressed accidentally.

    The 2086 and 3068 (Altos towers) moved the reset switch – too many people I guess had pressed it when wanting to extricate a floppy!

    Lots of memories coming back – thanks 🙂

  21. Mike Friese Says:

    I worked for Western Digital (WD), the company that supplied the disk controller for the on the Fortune 32:16. I designed the parallel section and firmware for the hard disk controller. Fortune’s controller was based on the WD’s WD1001 ECC disc controller that was intended for Seagate ST506 drives. It was based on five 20-pin WD-designed gate arrays and the Signetics 8X305 processor.

    The 8X305 was a special beast. It could execute instructions 3x faster than the Fortune 32:16’s native 68000 processor. But the 8X305 only had EIGHT instructions. Its speed allowed me to read 5Mbits/sec parallelized data off the disc and make real time decisions on that data. It also allowed me to implement the WD1001’s eight virtual host-facing registers in software.

    Since the Fortune 32:16 was all about performance, the disk controller needed to implement DMA. DMA controllers of the era typically handled only 8 bit data and 16 bit addressing. The 32:16 bus had 16 bit data and 24 bit addressing.

    I came up with a DMA solution that required no LSI devices nor counters. I had the 8X305 run the 24 bit counters and 8-to-16 bit bus conversion in software. The 8X300 would send addresses and data to simple, cheap octal latches.

    The Fortune engineers did not like this software solution because they perceived it as slow. Yet, in their own DMA specification, they required devices to not hog bus bandwidth.

    I demonstrated that the 8×305 software solution met both Fortune’s performance and non-bus hogging goals. The software loop required to update 16 bits of data, update the least significant byte of the address, maintain a word counter, initiate the DMA state machine, and wait for completion was only 8 8X305 instructions or 2 microseconds (us). The required 256 16-bit word DMA transfer could happen in just 512 uS. The sector time of the disk drive was about 1000 us. So my software DMA could deliver data to/from memory at full disc speed.

    (DMA transfers that crossed 8 bit and 16 bit boundaries took 2.75 us and 3.25 us, respectively.)

  22. Mark Bowles Says:

    I joined Fortune Systems in 1982, when it was located in the tip-up complex on Industrial Road in San Carlos.

    My specialty at the time was graphics. The graphics hardware for the 32:16 product line was designed and tested by Hub S; I developed platform software for it. Both of us were in-house employees, by the way, not contractors as others have reported. The initial products were to be a monochrome model, running on the same 12″ monitor as the primary console, and a color version offering 16 separate colors from a 16-bit RGB lookup table. Both systems offered 640×480 resolution. The color system drove a humongous CRT monitor, maybe 18″. A one-off prototype monitor and white plastic housing was prepared for our big announcement at COMDEX (maybe 1984?) — Hub flew it down to Vegas by purchasing it a separate airline seat. We called it the “white whale.”

    I developed a basic graphics platform software library for the 32:16, that provided a vocabulary of simple operations, e.g. move/draw, write text, set color maps, and so on. There was no standard available, and we had a COMDEX demo deadline to meet, so I simply built it to accommodate the demo I also prepared for the show — figuring that as long as I stuck to vanilla graphic operations, reasonably stated, we could retrofit it later to applications which the marketing team were lining up. I believe this is how this process went in the early days of the IBM PC — charting apps and so on were first built to run on specific graphic cards (e.g. Targa). It was only later that this Tower of Babel was simplified with standardized middleware.

    Coincidentally, this graphics platform also drove a small HP desktop plotter. This allowed our graphics dog-and-pony show at COMDEX to have a plotter quietly ticking away on a podium, drawing pie charts.

    For the monochrome system, I built a very simple Tektronix graphics terminal emulator, and found some canned files to run through it. These were complex engineering drawings and mathematical charts, that were very impressive in demo mode.

    The “white whale,” connected to a different 32:16, cranked out a continuous slide show of static images. As each image was being drawn in the background, the previous one was being displayed (using the lookup table to crossfade). One slide offered scanned black-and-white publicity photos of the three Fortune founders.

    Several visitors to the COMDEX booth accused us of cheating, and insisted on following the video cables from the “whale” back to the computer, convinced that there was a video player in the circuit somewhere. Fortune’s graphics offerings attracted a lot of attention at the show. The mission was to generate buzz from customers, but also to appeal to application software vendors, who the marketing team was actively courting.

    This was clearly an example of “demo engineering,” but it was in service of real product plans organized by Fortune’s marketing team. The monochrome version of the product did ship a few months later, but as Bass says, it was severely handicapped by a nearly total lack of application-level software. I believe my Tektronix emulator was one of the apps. I don’t think the color model was ever shipped, due to the expensive and bulky nature of the required monitor, and the lack of application software to make it useful.

    Some time during this period, Fortune moved to a sleek new building in Redwood Shores, near Marine World, adjacent to one of the ponds. Hub, the hardware engineer for the graphics system, was an expert wind-surfer, and taught a few of us how to surf during lunch hours, probably contributing to the heavy metal levels in all our bloodstreams. Eventually, he was so worn out that he took leave of absence from Fortune and became a wind-surfing instructor at a Club Med in the Caribbean.

    I left the company as an employee with the “no new development” announcement that John Bass mentioned in his 9/8/2018 posting. [Immediately after leaving, I spent eight months or so as a contractor to Fortune, designing the hardware and on-board software for a smart 6-port RS-232 card (which the Fortune marketing department imaginatively called the “COMM-6”)].

  23. Mark Bowles Says:

    John Bass mentioned Fortune’s software security scheme in his 9/8/2019 23:02 post. If we’re talking about the one used in the 32:16, I developed that in-house. It might have been after an outside vendor failed to deliver, but — memory fades.

    John is correct concerning the implementation details and the weaknesses of the scheme. It was used only to “protect” Fortune’s proprietary application software — primarily its word processing program. John, 10-bit machine ids does sound familiar. Hey, in those days, bits were expensive 🙂 .

    Distribution floppies provided the application code encrypted with a key known to the installation software. The floppies had a purposeful checksum error written into a sector unused by the unix file system. The installation software checked that the error was there, unbuttoned the application, read the machine id, and re-encrypted the application using the id. It then rewrote the erroneous sector. The operating system itself knew how to decrypt this version as it was being read into memory for execution.

    The objective of this system was to prevent “casual” pirating of the software — in particular, to force large organizations to actually purchase the applications for each machine that intended to use them. It would indeed be easy to hack, but only by deploying tools not ordinarily present in user environments. I think the company also wanted to preclude the “I didn’t know I was stealing the software, I just put the disk in the drive” defense.

    Unfortunately, the installation software was required to run on the smallest imaginable machine configuration (128k RAM, I think), so that the process of unbuttoning and re-buttoning the application had to proceed with very little buffering. This made it *painfully* slow. I think this may have led to some big customers insisting that the scheme be disabled for their purchases.

  24. Mark Bowles Says:

    Not enough has been said here about IMHO the major sales feature of the 32:16 — its word processing software, that was a purposeful clone of the Wang word processor.

    I believe that Wang’s fixed function word processors of the time were extremely popular among specific targeted markets, maybe including the legal profession. Fortune built a clone of the Wang word processing editor, and went to extraordinary lengths to make the replica as close to the original as possible. Since the Wang product ran only on its own proprietary terminals, and those terminals included special graphical characters, the Fortune terminals (both RS-232 and the hard-wired console) also provided these characters.

    At one of the trade shows I attended with Fortune, our booth was (hmm, coincidentally? I don’t think so :-)) located directly across the aisle from Wang’s. After an initial period of hostility, the sales people took turns crossing the aisle and giving their demos on their competitors’ hardware. (The show traffic was rather slow).

  25. John Bass Says:

    Hey Mark …. thanks for filling in missing pieces outside my own experiences )

    I completely agree about the Wang look alike comments … that drove our sales hard as an IBM Selectric replacement.

    And KUDO’s for the graphics and Com-6 work. Since you mentioned the in house graphics development, it did jog some very foggy memory’s … which included Rob playing with it in his office. And I remember the Tek emulator.

    I have one of the Com-6 cards that I got off ebay … awesome cool clean design. I don’t think I’ve ever seen one of the graphics cards. How many were made and sold? Were they primarily for one of the initial investor companies?

    I left Fortune not long after the Foster CIty move, nearly all my memories are in San Francisco and San Carlos periods.

    • Mark Bowles Says:

      The graphics module included a daughter board, and there were two versions: one for the monochrome output, and one for color. I doubt the color product ever went into production — its monitor spec was just too far ahead of the market. This memory is really vague, but I think the sole production run for the monochrome graphics board was around 50 units.

      I also imagine that the graphics module was more expensive than the product price could support. It had its own 8-bit 68000-family processor do do the bit-twiddling, and several PALs for driving the monitor.

      IMHO the demise of the 32:16 became inevitable with the introduction of the IBM PC/XT, in 1983. The fact that the PC line offered an open bus specification allowed third-party vendors to add value and make money off the machine’s success. In particular, this led to an explosion in graphics cards, and hence, graphics software that only ran on 8086/8088 architectures.

      Fortune moved to an Intel CPU just about the time I finished the COMM-6 contract, in early 1985, but IMHO this merely allowed the company to compete at a disadvantage to IBM itself, along with several other much bigger, much better financed clone manufacturers. The opportunity to provide significant product differentiation, in a market dominated by one product “standard,” had disappeared.

      • John Bass Says:

        During the 1981 to 1986 period the computer boom created several distinct core markets that were all successful. The real personal computer market for home use. The personal computer market for business desktop systems. And larger multi-user business systems, with central administration, which is where Fortune positioned itself in the market from the beginning. The single user PC market always had strong sales, even before Fortune, centered around a half dozen hardware/software standards … Apple, Z80 CPM, 8080 in several flavors including IBM. Business software and Hard drives were too expensive up until around 1987 to pay for per seat. The Intel move was driven by SCI after purchasing Fortune because that was the product line SCI engineers and manufacturing knew … SCI was IBM’s outsourced manufacturing arm for PC products, and lost that contract which sent them to buy a product line they could build a high volume business around.

        After Fortune I worked for a number of companies that positioned themselves in the same Multiuser UNIX market that Fortune created and dominated for several years. The best view of the market was the several years I spent at SCO supporting dozens of hardware platforms.

  26. Ben M. Laigo Says:

    I have a stock certificate….100 shares

  27. CitizenX Says:

    I was managing th Byte Shop Seattle store hen Lew had me check out a Fortune Mystems machine. It was a fine machine but I felt it didn’t fit into our retail focus. The had Alpha-Micros and North Star horizons which we didn’t have much success selling and I didn’t feel it was for us.

  28. Gary mahalak Says:

    My first job in the comp industry was working for a tiny Detroit area firm called Information technologies. My job was to fix,prep and install Fortune Systems 32:16 units. They eventually got them into Ford Motor in Dearborn. I still have my old Grey FprPro manual. great memories and glad to hear from you folks.

  29. John Leynes Says:

    You “old folks” may enjoy this 🙂
    This was the initial setup of computers in the North Florida Equipment Engineering Circuit Group in the new BellSouth Tower Building in Jacksonville, Florida. Honestly, at 73 now, I can’t remember the exact date, but it was around 1985…..
    See these on YT:
    The first is so “corny” it may also have been the first “Fun with Flags” or you could also call it “Fun with Fortunes”.
    BellSouth Fortune 32:16 UPS Demo
    Also: BERAM BellSouth Engineering Records And Markup
    I just found this site today and I don’t live here in the metaverse. If you like, you can contact me at, or leave a comment on YT.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: