Fable: Aspiring Expiring Architectures

In the 1990s two major computer companies of the time tried to get their in-house designed microprocessors adopted by outside companies in a bid to establish the processors as industry standards.

One of the processors was blazingly fast but the computer company which spawned it was flaky and collapsed and the part owning the processor was sold to Intel who killed off the device.

The other processor was sold in the computer company’s proprietary systems until 2008, but the computer company which owned it decided to throw its aspirations for an industry standard processor into a joint development venture with Intel for an x86 architecture device.


Moral: Intel omnia vincit




  1. I don’t think you have to wonder at all. Mike and Terrys’ Lawnmowers plc were well meaning and considerate employers but I don’t think they (corporately) had the knowledge or imagination to take the company where it needed to go.
    The UK VP’s finger was certainly on some kind of button. Lovely guy but he managed to elope with his secretary and disappeared off the face for about a month. Ho-hum, wrong sort of button methinks.
    The lack of direction was partly evidenced by the very rapid transition to Zarlink. The place became extremely factionalised with the management style invoked by a bastard hybrid of ex-Silicon Valley execs (failed) and an insanely devious British second tier. (Think Ed, the Serial CEO). Happy days – NOT.
    Anything not connected to MOD was sold and superb engineering teams splintered instead of being brought together. Shocking waste of talent. Half of them still driving taxis and touring in their campers.

  2. Interesting, Robtronics, one always wondered if Mitel Semiconductor really had its finger on the button of the semi business. it certainly had its finger on the button of British Telecom. I remember the disgust of Plessey Semis when BT insisted on having Mitel’s ISO-CMOS process adopted for the UK manufacturing of telecoms chips. To rub salt in the wound, Plessey had to pay Mitel £300,000 for the privilege. And Plessey could have had Toshiba’s CMOS process – the best CMOS process in the world at that time – for free.

  3. Working at Mitel in the early noughties we had the guys designing an 8bit SOC peripheral calling across from the next aisle,
    “Does anyone have a favourite instruction they want included?”
    Not kidding.

  4. SecretEuroPatentAgentMan

    Agreed, this is very interesting. And it makes me more curious, for instance how you decided on the instruction set architecture. ISA design seems to me to be an art more than an exact science, a topic that interests me. And what future developments may be in the pipeline?
    From the WDC web page:
    Annual volumes in the hundreds (100’s) of millions of units keep adding in a significant way to the estimated shipped volumes of five (5) to ten (10) billion units.
    Compared to 5 – 10 billion ARM processors sold annually the 6502 volume is impressive and interesting.

  5. Thanks Bill for an important piece of semiconductor industry history. Fascinating.

  6. The NMOS 6502 was design by a group of us who left Motorola on August 19, 1974 after designing the Motorola 6800. Rod Orgill and I completed the design of the Olivetti desk top CPU before the 6800 licensed by Motorola and manufactured and sold by Mostek as the 5065 MPU. So the 6501 (Rod’s project) and the 6502 (Bill’s project)where the third and fourth processors Rod and I designed. Chuck was the marketing guy with a great systems background and also wrote the 65xx software manual. The Western Design Center, Inc. I founded in 1978, I designed the W65C02S with the help of Larry Hittle. I also designed the 8/16-bit W65C816S. These processors powered the Apple IIc, Apple IIe and the ‘816 powered the Apple IIgs and the Super Nintendo. More history can be found here: http://www.WesternDesignCenter.com. Also, I was the designer of the original Mac mouse interface, the popular NMOS 6522 VIA, and replaced in the Mac with the W65C22S licensed and sourced by GTE, Rockwell, and others. WDC still sells these chips directly and through distributors like Jameco and Mouser. Cheers, -Bill

  7. SecretEuroPatentAgentMan

    Woz made SWEET16 to enable space efficient 16 bit capacity on an 8 bit processor (6502).
    Furber, Wilson et al made THUMB to enable space efficient 16 bit capacity on a 32 bit processor (ARM).
    Back in the day I was told that ARM was inspired philosophically by 6502 but I am not sure how true that is. SWEET16 is again supposedly inspired by an oddball CPU from 1975. Hmmm, do I feel another reader contributor quiz coming up?

  8. And the Alpha team went on to feed their clock distribution technologies into the StrongARM which kicked the ARM (or at least a compatible architecture) from around 33MHz to around 200MHz in one step.
    This allowed the manufacture of ARM based PDAs with sensible computer power for the first time, and was another factor in ARM success.

  9. Yes indeed, SecretEuroPatentMan, now that’s a thought.

  10. SecretEuroPatentAgentMan

    The world would have been a different place had they turned into hardware.

  11. Well, well, SWEET16, SecretEuroPatentMan.

  12. SecretEuroPatentAgentMan

    Not quite, though the 6502 reference is spot on. This paragon of elegant processor design powered the Apple 1 as well as the Apple 2 which was launched in the summer of 1977. The awesome designers of this marvel were Chuck Peddle and Bill Mensch (can you tell I really liked programming the 6502?) but Wozniak had no hand in the 6502 itself.
    On programming the Apple he found a need for 16 bit handling and made a virtual processor (or “soft processor”), implemented in 6502. This 16 bit processor followed much of the same design as the 6502 in terms of elegance, orthogonality and not needing prefix codes.
    One notable aspect is that the design was never implemented in silicon which is sad since it could have provided the logical path for 16 bit migration and perhaps even got there before the 68000.
    Now, surely, you can guess the name of the processor.

  13. Aha, SecretPatentEuroMan, so you’re talking about the 6502 plus the 2 256byte Proms for holding the virtual keyboard and the monitor control data, used in the Apple 1

  14. SecretEuroPatentAgentMan

    Not quite. 35 years ago is a big hint, let me clarify by naming the inventor, designer and implementer of this virtual processor: the Woz.

  15. I plump for the Ferranti F100l with second choice Natsemi’s PACE, SecretEuroPatentMan

  16. Apart from Digital’s (they didn’t want to be called DEC any longer by Alpha) bungling of the marketing and weird mix of other hardware (DECStations based on MIPs CPUs and by then obsolete VAXs) Alpha had two handicaps that I recall: no integer divide instruction so no modulo either, and the OS development was half-hearted to say the least.
    It was definitely a fast system in all its myriad variants, from mainframe-class down to PC-sized boxes.
    It’s also worth noting that one of the lead architects was Dirk Myer, who carried on the implementation of the EV7 System Bus in the Athlon CPUs and then the EV8 “direct connect architecture” in the Athlon64.

  17. SecretEuroPatentAgentMan

    I enjoy these trips on memory lane and discussions of by gone processors. However it seems that the anniversaries of a few have been overlooked, a little odd considering you listed one of them on a top 10 list.
    Allow me then to return the favour:
    35 years ago this summer a 16 bit processor was launched in an 8 bit world. This one however was simulated entirely in software, designed by a well known hardware and software wizard. What was the name of this processor?

  18. David, Mike,
    I did not use those either (they were anecdotal in EDA). From hearsay and wikipedia, alpha was faster than HPPA or Sparc by the end of the 90s for floating point and matrix, which does not necessarily translate into a snappy “user experience” on a workstation.

  19. That’s interesting Mike. I’ve never used a machine with an Alpha in it but I’ve always taken it as read that it’s claim to fame was sheer speed.

  20. Well my wife has used Alpha in the past and it’s very good for safety critical applications but was never the fastest processor around.

  21. Alpha, Mike

  22. Mike, I guess the alpha had more reputation than the HP-PA.
    In fact, I once heard a rumor that after HP acquired Compaq, they did not know what to do with the DEC alpha team, so they basically gave away the guys to intel, in exchange for some securing of the itanium, which HP for some reason wanted dearly for the next gen of servers.

  23. Which one is supposed to be ‘blazingly fast’?

  24. FYI in the late 90’s I was working for Philips who at the time tried the same with it’s until then proprietary TriMedia multi media processor. After that failed it got absorbed back in Philips and got killed almost a decade later.

  25. Absolutely, Stooriefit, spot on

  26. DEC’s Alpha and HP’s PA-RISC I think…
    The Itanium (which was/is HP’s JV with Intel) is like the old man in “Monty Python & the Holy Grail” who’s family want him taken off in the plague cart “I’m not dead yet! I’m getting better…”.
    I like this market forecast history:

Leave a Reply

Your email address will not be published. Required fields are marked *