It is a tale Told by an idiot, full of sound and fury, Signifying nothing.

SGI announced another RIF. Thats more good people tossed onto the street. Thats good technology about to be or already dumped.
SGI and I go way back, to 1993 when I started running molecular dynamics simulations on SGI machines. These machines were fast (those R3000’s were like butta …). I liked them so much, and liked playing with them so much that I joined the company straight out of graduate school.
There were many strange things about SGI. My introduction to the company was at a training function, and my first day at SGI was spent on a plane getting to that training. At this training function, I discovered that seniority was measured in weeks, that the senior execs, or at least some of them, had a frat house mentality, but were also insanely smart, or at least thats what we were told.
My first inkling of a problem was on the first day when the VP of marketing, a perfectly forgettable fellow who loved to name drop, allowed questions after his talk. I had the temerity to ask a question, using the phrase that *everyone* used to talk about SGI. That was, I called SGI, … SGI. Not Silicon Graphics. Just like all the other presenters. Just like this fellow did most of the time. Rather than answering my question, he was stuck on the “SGI” part. Couldn’t quite get past that.
I call that an “oh feces” moment, where you start questioning your decisions that led you to be at this location and at this time, facing an individual who so obviously needed to be fired (which he was later on), yet wound up in a position of authority in the company.
I saw the technology, and it was quite cool. But it was on a different trajectory than PC technology. It was on a slower trajectory, which meant it was going to be bypassed. I gave SGI’s mips technology about 4 years before it was passed up. Most of the technical folks I spoke with about this laughed. There was no way in their considered opinion that I could be correct, and they viewed it as naivete on my part.
Turns out I was wrong.
It was about 4.5 years before the PC technology integer performance surpassed that of MIPS, and about 6 years before the floating point performance followed suit. Nowadays, it is very hard to find MIPS based systems on the market. Well, no, thats wrong.
You can get them on eBay. I bought an O2 there for about 30$ US a few weeks ago. Call it nostalga.
The O2. One of the most awesome little boxes ever built. Could do things with media that are still hard to imagine PCs doing today.
Some of us were trying to push SGI management to sell the O2’s and IRIX very cheaply so as to increase the number of developers and therefore increase the number of applications. You have to do stuff like this if you sell platforms for applications. Otherwise you have this chicken and egg problem, and this is a problem who doesn’t have a solution you will like if you produce the chicken or the egg.
Management said no-go.
Some of us tried and temporarily succeeded at pushing SGI management to get into linux clusters. We thought that this market would be huge, and the downside to not being in it would be substantial. We were right. SGI got in, and then got out. In 2001, right before clusters took off. The IDC report summary I posted today indicated that cluster growth rates are in the 60-90% range per year. SGI does not have a product, 5 years later, that can compete in this space. It is quite likely that they will pull the classic augering-in move. Retreat to the midrange/high end.
Paraphrasing Abba Eban, they never missed an opportunity, to miss an opportunity.
There were one of the best places to work, with some of the brightest people, best hardware, best software. They could never get their message out. They could never understand why people liked challenge s units for web servers.
They made a video for a sales meeting in 1997. I found it recently, cleaning out some old stuff in drawers. This video had what I consider one of the best technical commercials for SGI’s strengths, and answering the question “why SGI”. It was funny. It was short and sweet. It was good.
This video never made it outside SGI.
I am sad to see them where they are. I departed after Warren Pratt made that fateful(fatal?) decision. They made many more blunders, not the least of which was believing a report that suggested that Intel would be shipping 10+M Itanium processors shortly after launch. I am not sure that they have shipped 1 M Itanium processors (including Itanium2) over the lifetime of the product. I have nothing against these machines/chips, we have an Itanium2 in our lab which is one of my favorite machines to work on. But SGI pinned its hopes on Intel. There was no plan B. Some of us were pushing for this plan B to at least exist.
I am sad to see them in this state. SGI gave me many fond memories, with excellent colleagues, wonderful customers. But there are many painful memories of poorly thought out decisions, poor execution, purchasing a competitor we were previously busily destroying… lots of dumb things.
Sometimes dumb things, or enough of them in a row prove fatal. Death by a thousand paper cuts. Or enough really bad decisions.

1 thought on “It is a tale Told by an idiot, full of sound and fury, Signifying nothing.”

  1. It’s too bad what’s been happening to SGI. First they buy out Cray, then let them go, then decide to enter the supercomputing market on their own. They also ruled the server market at the beginning of the dot-com boom, only to loose that to Sun. And then there’s the graphics sector, which became synonymous with Apple long before Macs were ready for the task. I hope their new CEO can turn things around, though I suspect it will really just collapse (or be taken over by HP out of pity).

Comments are closed.