The Supermen: The Story of Seymour Cray and the Technical Wizards Behind the Supercomputer Book

Seymour Cray is one of the greatest virtuoso engineers of all time.

I'd read about him in various other books. He was noted for his eccentricities. Instead of working in some lab in a major city, he worked in a cottage overlooking a lake in Wisconsin (Lake Wissota, Chippewa Falls, northeast of Eau Claire, roughly 90 miles east of Minneapolis). He'd build a sailboat, sail it on the lake and, next spring, burn that one and start building another. Each one better than the last.

The "larger" story is a bit more complex than that. And this book is largely about him. Sure, it begins with other people, working for the Navy during WW II. But he becomes an important figure pretty quickly and the book ends with his passing in 1996.

I never met him but I was aware of him. I was a "computer geek" in high school and, among computer geeks, the name "Cray" was spoken with reverence. He made the fastest computers in the world. That's not an exaggeration.

This story isn't so much about the computers. Some technical specs are given, but not that much. I was a little disappointed in that regard; 4 stars instead of 5 'cuz I would've loved to "geek out" over more of the details. But his biggest struggles in life sound oddly familiar in my own life.

I'm a programmer or, in more modern parlance, a "software engineer." An engineer takes known principles of science and applies them to create products. Engineers may get involved in research, but that's usually a means to an end; we need to create this product, and some of the details aren't fully understood, so we'll do some research to finish fleshing it out. Scientists focus on research, engineers on the application of what they discover. This is true for electrical / hardware / computer engineers, like Cray, and for software engineers like yours truly.

The problem for people like me is that, beyond a certain point, companies don't seem to have much use for us. We progress to a certain level and ... if you want to make more money, you have to go into management. The management skillset is very different from the engineering skillset. Insomuch as I've dedicated multiple years to getting a degree and decades gaining experience at what I do, why would I want to set aside all that training and experience and start, from scratch, on a new skillset?

Theoretically, a knowledgeable engineer knows better than anyone how to manage other engineers. But theory and practice ... not the same thing. The Peter Principle states that everyone rises to the level of their incompetence. If you're good at what you do, you move up. When you're no longer good at what you do, that's where you stop. Why would any, intelligently-run organization want to take people who are very competent at one job and move them up until they hit a level they're not good at? And yet, that's how it works.

Then, as now, companies really don't know what to do with experienced engineers who just want to be engineers. Some of them genuinely try to create a career path where people with significant technical skills and experience get to use it, designing stuff for others to build and mentoring the less-experienced. But for all to many ... the only way up is through management. Which engineers frequently see as "crossing to the Dark Side."

A common refrain in this book is that, when things got too "corporatized," the engineers left and started a new company. One where they were needed as engineers. One where it was clearly understood that the company lived / died based on just how good they were, just how hard they worked, as engineers. Management was kept to a minimum.

It starts with Engineering Research Associates (ERA). They got bought and the engineers chafed at the management of their parent company. So they left and founded Control Data Corporation (CDC). When Cray found himself spending more time shaking hands and reporting to management, he became what is more commonly known (today) as an "intrapreneur." Without spinning off a separate company (that would be an "entrepreneur"), he took his engineering team, moved them away from the all the corporate distractions (hello, Chippewa Falls) and then turned around and delivered. They were getting nowhere in the old environment; in the new environment, engineers were free to be engineers and the management was far away. So they succeeded.

Not "and they succeeded." "So they succeeded." There is implied causation between separation and success. And any engineer quickly discovers that there is a definite causative relationship between micromanagement and failure.

When CDC's management became too much, Cray became a contractor. Again, things got too "corporatized" and he left, forming Cray Research and introducing the Cray 1. The CDC machines are little before my time but the Cray 1 ... ooh yeah, I remember those. I have pictures of me posing with one at the Computer History Museum. For me, actually getting to touch a Cray 1 was a Bucket List item.

The follow-up, the Cray 2? Oh yeah, I remember that one too. There's never been another computer like it, before or since, that I'm aware. I mean, c'mon: that thing immerses the electronics in liquid (?!) and uses a FOUNTAIN (!!) to cool the liquid. No, I'm not joking. There were practical reasons for that design.

No one designed machines like Seymour Cray. No. One.

When Cray Research became too "corporatized," he left. Again. Unfortunately, the next venture was not successful.

As with so many small tech companies, there's frequently a race: can we finish a successful product before we run out of money? With CDC and Cray Research, success arrived first. For Steve Chen's Supercomputer Systems Incorporated (SSI; Steve Chen is sometimes regarded as the designer of two Cray systems, the X-MP and Y-MP) and for Cray Computer Company (CCC; Seymour's next act after Cray Research) ... the money ran out first. For each of those companies, there were some very interesting technologies being developed.

SSI was developing a system to cool the chips which involved spraying coolant on them; the coolant had a low boiling point and flashed to "steam" on contact, carrying away the heat (phase-change cooling is EXTREMELY effective). They were working on multi-chip carriers, meaning one ceramic casing had not one but multiple chips inside; by cramming the individual chips closer together than the single-chip packaging (still common, today) allowed, they could push speeds even higher. They were also working on multi-ported RAM (not just dual-ported, as found in many video cards; and this was LONG before video cards existed) and light-based timing, instead of electrical (the norm, still today).

For CCC, Cray had decided that he'd ridden silicon as far as it could go and was manufacturing chips based on gallium arsenide, which was 6x as fast as comparable silicon chips. He was also tightly stacking the component boards onto solid blocks, with circuit pathways going in all 3 dimensions.

Neither of those companies survived and, so far as I'm aware, neither of those technologies did, either. I can only imagine how fast PCs and tablets would be, today, if they were using gallium arsenide instead of silicon. As for multi-chip carriers, modern smartphone CPUs are actually microcontrollers, integrating multiple functions into a single chip. Instead of cramming multiple chips into one package, we've crammed the equivalent of several chips into a single chip and stuffed THAT into one package.

In both cases (SSI and CCC), the founders were trying to create ideal environments for engineers to be engineers, with a minimum of "corporatization." Sure, this engineer is more of a leader / mentor to other, junior engineers. But few, if any, are purely "management;" almost everyone "works for a living." Unfortunately, success (especially if you go public) means you will need some dedicated management; no one seems to have found a way to automate away all the management (a very worthwhile goal, IMHO). Once you get too many management and ancillary positions, it's only a matter of time before people forget who is creating the products that create the wealth. And when those people are unappreciated ... they leave. After which it's only a matter of time before the wealth is no longer created and it all goes downhill (see CDC and Cray Research). The only alternative, from management's point of view, is to view the engineers as disposable / replaceable. When Seymour Cray left Cray Research, they had to have a new "genius" designer; when Seymour left, investors left. Hence Steve Chen being promoted as the "creator" of the X-MP and Y-MP (yes, he contributed, but both were variations on the Cray 1 and largely orchestrated by another engineer who didn't want the limelight). After a well-run publicity campaign, the investors calmed down because, although Seymour was gone, there was a new "genius" at the helm.

Doesn't that sound all too familiar?

I got really tired of all the corporate mess, myself. I'm kinda like Seymour Cray in his Chippewa Falls facility. I go to the office very few days. I'm in contact with those who care, and I can communicate effectively without needing to be there, physically (99+% of the time, anyway). Reporters noted that the Chippewa Falls lab had plenty of windows, set next to a forested area. There was a salt lick just outside the front door, so deer would frequent the area. I have a nice, wide window looking out over grass and a forested area. Both of these are examples of Christopher Alexander's "Windows Overlooking Life" pattern, noted for helping to spur creativity. Seymour's "commute" was a couple hundred yards, on a path through the woods from his house to his lab. Mine is within my house.

I totally empathize with what he was trying to accomplish. I'm nowhere near his level of virtuosity in my craft but ... I can understand. He just wanted to be the best engineer he could be.