The Adventure of the Suicidal IC's

With ICs jumping off boards and more ready to make a break for it, an engineer helps out a competitor

By Bob Cowell, Contributing Writer

Around 1982, the new hotshot Silicon Valley company was Silicon Graphics (SGI), rumored to have a killer computer graphics demo. I was still at Carnegie-Mellon University as a grad student, but was working part-time at Three Rivers Computer, an engineering workstation startup that was paying me to develop a color graphics version of their original black-and-white workstation.

I had gotten involved with Three Rivers (later renamed to Perq Systems) for two reasons: 1) I had seen a really fascinating demo by them with graphical windows sliding around on the screen, under and over each other, and I simply had to know how they were doing that (turns out they were faking it, at least originally); and 2) I wanted to design things as a break from grad school research. Because I was designing a color graphics engine, I was eager to see how my design would stack up against the new kid on the block.

I arrived early at the classroom where the SGI demo was to be held, only to find the SGI presenter muttering to himself with his head stuck in the back of the chassis. It quickly became clear that his demo machine was not working. Argh. I really wanted to see that demo.

Several minutes later it was becoming clear that this demo was not going to happen. So I figured I might as well see what was in the chassis; at least this presentation wouldn’t be a complete waste of time. I asked the presenter if he minded if I looked around, and with an exasperated sigh he said sure.

The first thing I noticed were the usual array of printed circuit cards stacked on their sides, plugged into a backplane. Nothing unusual there; looked like lots of TTL. But what DID seem unusual was that there were approximately ten IC’s lying on their backs on the bottom of the chassis.

Now, that was a problem I knew something about - the old DIP IC’s often had a tendency to “walk” out of their sockets due to the vibration of the cooling fans or during transportation. Some socket types were far worse than others for this problem. I sometimes resorted to using little nylon cable clamp “seat-belts” to keep the IC’s from going on walkabout.

I asked the presenter if he’d mind if I tried to find out where these IC’s had fallen out. He probably figured I couldn’t really do any harm, and waved a hand in my direction that I took to mean “okay.” It’s not as hard to find the proper homes for these IC’s as you might think, because many of them are “sliced”; a ‘244′ octal buffer is likely to be found in the company of other ‘244’s to carry a 32-bit bus, for instance.

In about ten minutes I had found likely homes for all of the escaped IC’s, re-socketed some other chips that were “making a break for it,” and suggested to the presenter that he try the machine again. I held my breath - it worked. And the darned demo was great, to the point where I had to mentally revise the expected market for what I was designing. I really enjoyed seeing SGI’s demo, but I always wondered if directly helping a competitor was a legitimate part of my job description.

Contributing Writer Bob Colwell was Intel’s chief x86 architect in the 1990s and has worked as a computer designer at VLIW pioneer Multiflow, Perq Systems, and Bell Labs. Author of The Pentium Chronicles and the At Random column in Computer Magazine 2002-2005. He is currently an independent consultant.