Four Early Computers, 3 and 4

This is a continuation of Four Early Computers: 1 and 2. The story opens as our intrepid writer...that would be me...is still in the middle of his final year as an undergraduate studying computer science.

Data General Nova

Another computer we used on the CS course was a Data General Nova. I don't know the serial number of ours, they have #1 in the CHM. It was one of the 16-bit so-called minicomputers. Over 50,000 of them were shipped. Most operations took about 2.5us so I suppose we would say it ran at 1/4 MIPS. It had no virtual memory, nor memory protection. However, it did have an operating system called RDOS (real-time disk operating system).

This was the computer that we used for learning assembly language programming. It was also a computer you could do what you wanted with since you were the only user. Like most minicomputers in that era, if you somehow overwrote the bootstrap that would reload the operating system, you would have to key it in binary from the front panel switches you can see in the above picture.

Data General was very successful with this machine and decided to build a 32-bit version. The design of this is immortalized in another book, Tracy Kidder's The Soul of a New Machine. The only book that comes close to capturing the realities of high-pressure engineering projects is David Lundstrom's A Few Good Men from Univac, about the transition from engineering to politics and business, and especially the career of Seymour Cray.

Vax 11/780

Much of the programming when I got to Edinburgh University to do my PhD in CS was done on another computer unrepresented in the CHM, the Interdata. These were 16-bit minicomputers with an instruction set roughly based on the IBM 360. We had over a dozen of these, used for undergraduate teaching mainly. My research was on distributed file systems and I built a network file system using one, and then a standalone diskless operating system called Legos.

The CS department also used the university computing service mainframes, which ran on ICL 2900 and a forerunner whose number I forget. But we decided we wanted to be self-sufficient and when Digital Equipment Corporation (DEC) released the Vax 11/780, basically a 32-bit version of the PDP11, it seemed a good fit. It was ordered, along with some disks and a disk controller from a company called System Industries (SI). This turned out to be a fiasco, since System Industries had no device drivers for Vax. Their bread and butter had been selling to PDP 9 customers since their solution was a lot cheaper than that from DEC. By then I was an employee of the university, as well as working towards my PhD, with the grandiose title of "computing officer".

In a brilliant bit of social engineering, my PhD supervisor, who was also the person in the CS department responsible for hardware acquisition and so had purchased the disks, challenged me that I couldn't get the SI disks working by writing a device driver for them. The documentation for writing a device driver was all but non-existent, but the saving grace was that the entire source code of VMS, the Vax operating system, was on microfiche. In some ways it was already "open source."

It was all written in assembly code, but the device drivers for all of DEC's own devices were there so I could learn by example. Soon, I had them working, but you still needed to boot the operating system off the smaller, slower DEC disk and so the system disk, used for paging, was slow. The next challenge was to boot VMS from a non-DEC disk. This turned out to be simpler than I expected. The Vax had a control processor that had a floppy disk, and the control processor would read the bootstrap from the floppy disk and load it into memory, and then run it. So I created a version of the bootstrap that used the SI disks and it worked.

We had the only Vax in the world that ran on non-DEC disks, and we had the only Vax in the world that used SI disks at all. In approximately five microseconds, I was on a plane to Silicon Valley for my first trip to the US since SI were on Oakmead Parkway in Sunnyvale. Since I was being paid by the University, it never occurred to me or anyone else to ask for a consulting fee, so I basically taught SI how to get their disks working for free. They didn't actually even have a Vax but Lawrence Berkeley Laboratory had a Vax and had also bought SI disks that didn't work, so they were happy to let us use it since it was useless until the disks worked.

The next challenge was that our Vax was isolated from the university-wide data network since we didn't have drivers for the networking hardware nor a network stack for the Vax. The university computing service would work on it, but they said it would take a year or two. Another PhD student called Gordon Brebner and I had two weeks of downtime during Easter vacation (he would go on to be the head of the Department of Informatics at Edinburgh, before coming to the US and working in research for Xilinx). At the end of two weeks it was working.

The Vax 11/780 became the workhorse of Silicon Valley and EDA. In fact Gordon Bell had an aphorism that all business plans start "Raise Money, Buy a Vax...". Then, like the South Park underpants gnomes, something happens, "...profit."

Today

Obviously, without these computers I wouldn't be doing what I am today or probably even living in California. Almost anyone who has been in the computing industry or related areas for a length of time will find key computers of their life in the Computer History Museum. Maybe a Sun workstation. Maybe a Cray 1. Maybe just an early PC.

Anyway, to finish the story, I finished my PhD and interviewed at several companies in the US: Bell Labs, DEC, HP, Intel...and a little startup called VLSI Technology. That day I learnt a lesson about hiring that I have never forgotten in my entire career. At 4pm on the day they interviewed me, they gave me an offer letter. I'd actually interviewed at the other companies earlier, VLSI was my final interview. Those companies told me they were working on offers, but they were all stuck in their HR bureaucracies. If you decide you want to hire someone, give them an offer letter already.

Since my vague plan was to come to the US for just a couple of years—Silicon Valley experience would look good on my resume back in Britain—it didn't much matter where I worked and a small company seemed interesting. The CFO had to sit down with me and explain what stock was, since they were giving me some and I didn't have a clue (he was Ken Goldman who would be the CFO of Sybase, Siebel, and currently Yahoo). I accepted my offer. I worked there for 16 years, and I left because I was CEO of the Compass Design Automation subsidiary and sawed off the branch I was sitting on by orchestrating its acquisition by Avant!

I worked at Avant! for eight hours. We had to do a Hart-Scott-Rodino filing (don't ask), which has a 30-day waiting period. That ended as Thursday moved to Friday. I resigned on Friday thus earning my bonus for a successful sale. But in the meantime I'd been interviewing. First thing on Monday morning I started as the VP Engineering at Ambit Design Systems. A year after that Cadence acquired us, and I started my first tour of duty here.

That two-year plan...well, I'm still here...although I did live in France for six years in the middle. But I've never needed a good resume in Britain since.

The Circle of Life

So computers used to be big unwieldy things that filled entire rooms. Then they got to be only the size of refrigerators, then small dorm-room refrigerators.

Maurice Wilkes was the head of the Cambridge Computer Laboratory when I was there. I remember someone challenging him in about 1975 that due to speed of light considerations, computers couldn't get any faster. In those days it was normal for a mainframe processor to be on one side of the room in a big box, and the memory maybe 10 feet across the room in another big box. Wilkes thought for a moment and replied, "I think computers are going to get a lot smaller."

And they did. The microprocessor was invented at Intel. They have their own museum, too. That was the first one I wrote about in The Intel Museum.