The Sad Saga of Silicon Graphics: The Final Chapter

I worked for SGI in the late 1990's. It was obvious even then that they were making a fatal mistake. As other server prices fell, instead of trying to lower costs, they tried to squeeze more revenue out of their installed base. They raised prices for support and training, and drove customers away. Of course, focusing on high-end systems when intel processors were running rings around SGIs proprietary MIPS chip didn't help either. Sigh.

I know someone who use to work for SGI back in its heyday. He told me that he had some friends there that were worth billions (on paper) due to their stock options. He met some of them for lunch several years later and now these guys were worth "only" millions due to the decline in the stock price. I feel bad for those guys knowing that today they are probably worth far less than that. Too bad, because SGI had some great technology and some really smart people before the crash.

I some ways it's nice to see the multitude of former SGI buildings re-occupied by the next generation tech company: Google.

In the early 90's some noticed a trend: company's that built new fancy HQ buildings soon floundered.. Amdahl, SGI, AMD, Yahoo, to name a few. Might it have been excessive exuberance on a micro-tech scale?

This is a sad day. In the early 1990's SGI computers shaped my career path. I went to Temple University and we had 3 SGI's, none in the film and video department. I connected with an SGI reseller who eventually became my employer.

We sold SGI systems running Softimage 3D. At the time a mid-level SGI could run circles around any other platform on the market and as a reseller the margins were very healthy.

I even sold and deployed an SGI 3D animation lab in Hanoi, Vietnam and spent a month there teaching the customer how to use the platform.

Some other things I will also always remember was SGI's arrogance by thinking that no other computer maker could touch them. They so heavily ensconed themselves in their own proprietary hardware and software that they could care less. That was 1994.

Proof of commoditization: In 1994 a mid-range SGI system sold for $55K and up. Today a 2006 or 2007 SGI system that probably sold new for $20K can be had on eBay for under $200.00.

They put all of their eggs in one basket and the bottom fell out very slowly over the last 13 years.

With companies like Infiscale giving away software that turns commodity gear into better than SGI performance, what do you expect?

Why so glum chum? Rackable Systems has mentioned that they want to get into the HPC space by buying SGI. I still have hope that innovation from SGI will rein again... only now it will be called Rackable Systems.

Fail - end of story

I may be one of the only people to have been part of a company collapse that was actually bigger then SGI, and also to have been part of SGI as well. I worked for Control Data for ten years, then for SGI the next four years. Control Data went from $4Billion in the 70s to zero. SGI was actually less of a flop in inflation dollars. Both were innovative companies that dominated a segment of the technical computing market for a significant time. Both fell victim to the onslaught of the commodity computing growth. Volume wins over innovation every time. Short term victory can be achieved by brilliant engineers, but the volume of a commodity computing product will eventually win the market share, the performance game, as well as the price point competition.

I learned a lot at each company and each had a great run, but they and I had to move on to find the next opportunity. We both were more, or less, successful at this endeavor.

RIP SGI and CDC

SGI lost vision years ago. This has been over a decade in the making. Its roots are in the early 1990s. And they could have avoided it if they had only listened to some of their more creative employees. They had the potential to do some amazing things but they just blew it. Their original graphics market was eroded by companies like Sun from below and then when Wintel computers had top of the line graphics driven by the gaming industry. They failed to diversify soon enough. Their purchase and subsequent sale of Cray was insane and even so they could have capitalized it if they focused that technology on database storage instead of visualization. Imagine a database server running on a full blown Cray supercomputer back then. Instead, Sun went on to revitalize itself temporarily doing exactly that with Cray technology that SGI actually gave to them because they didn't want it. IBM is flying high right now but they should take a lesson from this. How the mighty have fallen. Of course SGI was selling for about .02 a share a few years ago so it's really rather remarkable that they lasted this long. IBM is a lot more diversified and if they buy Sun I guess they get Java which is the essence of WebSphere so this is exactly what I'm talking about that SGI failed to capitalize on when they had the chance. They just blew one opportunity after another. Even Barbie has competition. And you can bet Mattel is looking sharp at the Bratz. Ken is probably already sleeping with them.

Id say SGI was still a stepping stone in what technology is today and I see the workers of SGI historically.

Netap is another company that makes overly expensive systems using proprietary software, they are threatened by open source software on commodity hardware. There's little wonder then that they are in a patent lawsuit with Sun. The only trouble is that Sun is about to be acquired by IBM, which has an even bigger patent portfolio than Sun, and what's worse is that IBM is a customer of Netap.
Eric

Bigger falls? DEC, DG, Wang, CDC . . .

Too bad. I remember while attending the NAB (National Association of Broadcasters) shows in the late 90's, Silicon Graphics booth with it's amazing theater was always the "can't wait to see what they are doing this year" destination at the show.

Your column is titled "tech beat", but I fail to find a tab/link on this page to help me share the story on popular blogging/social networking sites.
Try to keep up with the tech!!! Good article otherwise.

Louise, DEC, DG and WANG all sold for more than $25 million and I don't think any of them went into bankruptcy so I can't see how they count as bigger falls.

CDC I don't know too much about, other than they never really got big enough to fall that far...

Sometimes, no matter how far ahead you may be in this field, if you don't address the overall commercial needs of enough commercial customers, you just can't make enough money to beat your expenses, particularly when the bottom falls out on continuing credit and when taking on R&D and other projects with skyrocketing expense. No doubt, SGI tried very hard and it also wasted a lot of its own momentum and got a bit strange at times. We will all miss them and hope under RS they will develop distinctiveness, even as a hedge they could endup taking over the future of RS. An enormously valuable $25 million investment for RS.

I came from a Unix background into Windowed Systems from before and during the Microsoft era as a systems developer (hw and sw) and an early GKS and Video on Computer developer. One day, many years later, working at Sarnoff on a visual project on Sun, I noted that despite the speed of the integration of RISC with OpenGL hardware, I came to believe with SGI I was looking at was eventually going to be thought of not as a mass provider but as a small volume off prototyping company whose products would see limited use and limited backing from the mainstream. Others would benefit from SGI's innovations as they proliferated into the industry, a sort of Wang like company which eventually collapsed and left a legacy behind, pardon the comparison with Dr. Wangs Lab...

That has for now turned out to be true. We all get carried away with belief in a particular solution. Whether it be MIPS, or OpenGL or 3G Glam, no matter what it is, we just don't seem to learn that "those with the gold make the rules" and unless you can do a Google, building bricks and mortar and putting your logo on it is usually self defeating.

But Bob Pearson, you were at both CDC and SGI? Yes, you are correct, they seem to be similar kinds of collapses. Well, they have one thing in common (You!!!). Just kidding Pearson... They had other things in common. You know that big company in Maynard? And those wealthy New York family? And that BOE Telephone Company?

End of an era, Bob, sometimes means rebirth in legacy. We can all hope.

Maybe IBM will buy NetApp too. Solaris, NetApp, whatever, it will all be called either Tivoli or WebSphere. :-)

innovators dilemma. read it. sums it up nicely.

Usually what happens at these companies is that management loses touch with reality, focuses on the glory and the money and themselves, and forgets about the employees who were actually the ones supplying the majority of the creativity and energy that made the company what it is. The employees then go into misdirection, many leave, creativity and know how is lost, motivation levels are lowered. At that point, the company starts to unravel and its the beginning of the end. This common chain of events usually never gets observed from outside the company and that story is never told.

How the mighty have fallen. As a young kid just getting into 3D I still remember drooling over the million dollar SGI dream machines of the early 90s. Those three letters still invoke feelings of wondrous out-of-reach technology for me. I never did own one. 3D software got ported to Windows and that was it.

It's really too bad about SGI. I remember when I worked as a 3D artist back in the mid 90s, SGI's were pinnacle of graphics development. Then I was able to do the same thing with a far less expensive PC and eventually an inexpensive laptop running all open source software. The bottom line is that aside from corporate mismanagement, they were arrogant and their reps considered themselves untouchable--I remember them, they didn't have time for customer service. The simply didn't think that the competition would ever surpass them. Good bye SGI, there's a bunch of little purple boxes on eBay now for next to nothing and I won't buy one because now, my cheap laptop is faster than an SGI Indy--something that I had always wanted...

Kinda sad to see SGI go. Wonder if Rackable will change to SGI or not.

Jim, Mattel just sued MGA and are now the owners of the Bratz line... so even with Mattel came out ahead... not so lucky for SGI.

My company Digital Art was the first int. buyer of & 2400 Turbo SGI and I became the exclusive distributor for SGI in Europe in 1983. They were the ultimate graphics and 3D workstations on the world market. For each machine we bought we needed to get an export license from State dept, since it was regarded as strategic equipment. Their tech support for us was absolutely minimal, we needed to stock spare parts (obsolete in a couple of months). In my offices we had 17 workstations lined up for simulation and 3D animation development. With SGI and Pixar we became a leading image synthesis company in Europe. Two guys I really liked working with: Jim Clark and Steve jobs, they kind of people who knew what needed to be done. A great 10 years of innovation I was part of.

From a superstitious outlook, SGI's downward tumble accelerated when they changed their logo. Coincidence? Nothing is!

We used SGI workstations at school in the mid-90's, along with Suns, HP's, and DEC's. It was never clear why they were better, except they were the newest, and had the nicest desktop backgrounds. This was just before Wintel days, so SGI and Sun both had "workstation" cred, which faded quickly when machines got dirt cheap.

I also used SGI's and an Origin 4-blade server to do a film project; the Origin was supposedly the fastest machine on campus, but it still took 2 days to render, and nowadays common Wintel desktops are quad-core anyway.

VRML was a promising venture, but way ahead of its time in 1994. If SGI had 10-25 developers and VRML, they might actually be a viable company today.

DEC had more potential than all of them in my opinion, particularly with the Alpha processor...they were way ahead of everybody at the time, even when they declared bankrupcy and sold off the pieces to Intel, Compaq, Samsung, et al... Binary compatibility with x86 always kills them in the end when they try to compete directly with the Intel franchise...

I think perhaps Intel could be in hot water this time around since they have igored the low-end of the market where ARM has become the de factor standard for portable devices ... now the portable devices are moving rapidly up the food chain like little piranhas chopping away on Intel's lunch eventually...ipods, blackberries, etc... next think you know iPhone is a full fledged PC computer with half the power requirements of a power hungery Intel core...crazy perhaps... but certainly possible... more likely than Sun's Java workstations conquering Windows and Microsoft... really wondering what they were smoking in the late 90's... :-D

There are few better examples of being on the wrong side of disruptive technology. This will amputate the last bits legacy tech and perhaps, as stated earlier, the inovation of SGI will live on.

years ago i lusted after SGI workstations - any workstation at any price - but could never afford one until now - and now i don't want one because it's old technology!

So many of you say "It is a big mistake not to jump to Wintel bandwagon." And "Volume beats high techology."

It is the same that making Leonardo to carve chairlegs. It it so hard to give up with decadelong developments resulting high standard products against cheapo thingabobs wooing the non-IT masses.

Oh yes. Leonardo lies in a grave, his paintings and statues eroding, but we still sit on chairs.

I don't know what type "pre-packaged bankruptcy similar to the ones being considered for GM and Chrysler right now" that" "allows operation in shareholders interest" that the author is talking about. SGI did go through bankruptcy several years ago, and as a share holder back then...I lost every penny.

Plus what about the separate company (arm of sgi) that does work for the military and government.....and or the German and Japanese companies... They were not part of SGI as I understood it.

Good article - needs a better title. How about "Indigo Fades to Black"?

It takes resources (money and time) to innovate. Commodity tech tends to reduce resources available for innovation in the race to the bottom ( in cost). If it wasn't for these large tech companies competing with their tech, we might not have the computer HW we have today. So, we all appreciate their innovation over time as it becomes more affordable when integrated into commodity tech that was once high end proprietary tech. Many of the people and Ideas they generate will go on to other companies to benefit us all eventually. Long live those once successful but now dead And absorbed or constantly floundaring high end tech companies!

SGI made terrible equipment, and Stanford had no use for it. It ended up buying from other vendors.

I worked with (not for) SGI in the 90's. The culture there was the epitome of hubris personified as its employees largely thought that they were so good that they could be arrogant asses.

In the end, they had an advantage that they let slip away - SGI made a conscious, strategic decision to abandon the render/process advantages of using 3D Constructive Solid Geometry and threw their then industry-dictating weight behind Polygonal Modeling which eventually spelled their doom.

Once they lowered that core visual standard, it was only a matter of time until the ability to throw render mountains of cheap WINTEL boxes at melding polygons would undermine SGI's business model. We at electroGIG told them this and they scoffed.

well, it doesn't feel good to say "told you so" when it dragged down both SGI and a few other good companies... as well as condemning the world to largely unrealistic 3D effects and virtual sets but there it is.

SGI RIP

Interesting reading the comments from people from CDC. I did three years as a contractor for CDC's timesharing business back in the 80's. Told them at the time that PC's would eat their lunch - nobody listened. Sounds like SGI did about the same - market switched, they didn't. I guess a couple of lessons from these busts is that just because something worked yesterday doesn't mean it'll work tomorrow, and money flows downhill, i.e the cheapest adequate solution to a problem wins.

I consider myself fortunate and blessed for being allowed to be apart of the SGI culture in the 90's. We worked extemely hard, we played hard, we had great co-worker camaraderie, and our customers & our competitors desired to be part of that history. My only regret...is that it didn't last till retirement. I miss it. Thx SGI!

Nothing sad here really, things change. Rackable (aka smackable) will probably gain - at the very least - a sense of style.

Like many have summarized above, the only surprise is how long they've lasted. I was one of the guys on the outside, trying to get in around 1994 and was shocked by the total arrogance of the organization at all levels. They had a lot to be proud of, but they let it go to their heads.

They could move even today to offering high end Intel multi-cpu workstations but they didn't.

SGI's DNA was built around the brilliant visionary leader Jim Clark, exceptional talent, a hybrid academic (Stanford)/commercial culture, a super work hard/play hard environment, and track record of giving customers access to technologies that few other companies in the world could dream of, much less deliver. SGI has had a much more positive impact and influence in Silicon Valley than many people realize. It was, for its time, a great company.

In the nineties my company was a customer of SGI. We bought SGI servers to do high end modeling. SGI worked very closely in deploying servers at several locations in our company at no initial cost to my company. This was just to prove the business value of such high end servers. We proved the business value and bought more servers from SGI. It was such a pleasure to work with SGI personnel to apply high end modeling concepts in different areas.

Then the paradigm shift took place. High end modeling which was possible only SGI and Evans & Sutherland servers/stations, was now possible on Apple Macintosh and PCs. SGI lost its bearing. The head-in-sand mindset was so mind boggling to me. The same company which was so innovative and customer focused, now was not able to see the writing on the wall. SGI did not make the shift and started digging their own hole.

SGI is a great business case that all aspiring MBA students should study carefully. There are lots of lessons to be learned from this experience. I was a first hand witness to see "frog in a boiling water" syndrome play out with SGI right in front of my eyes!

Adios SGI

I used to deliver training sims back in the mid to late-90s that heavily leveraged SGIs, initially for graphics and later for both host function, physics, and graphics. To slag them off as not being the best of the best back in that era is to deny reality (or perhaps Reality Engines, if anyone remembers that technology line...).

From where I sit, in 1995 or so they were not just performance superior to their competition, but cost competitive as well. Graphics wise, the only thing truly equal (and maybe even superior in some settings) from a performance standpoint, were the Evans and Sutherland products. But these were a completely proprietary closed solution, ill-suited to any function other than throwing pixels. And at costs per graphics channel an order of magnitude more than a comparable level of resolution SGI system. I could (and regularly did) put SGI racks to work running computational fluid dynamics and finite element models of ridiculous complexity overnight, then analyzed results, and retasked them to throw pixels in the training sims all day. Try that with an E&S... or anything else for that matter.

Earlier posters noted, and rightfully so, that SGI completely missed the boat with the growth of Wintel and anti-aliasing PC graphics cards. I vividly remember the way their first Wintel boxes, running NT4, were offered, but held at arms length with pinched noses, back in the late 90s. Looked down upon from within and poorly pushed without, it is not surprising that SGI failed to see they threat welling up underneath them and react accordingly, especially given their self-absorbed and egoistic world view. Yes, their arrogance was legendary. Of course, the same folks slagging them off for that arrogance probably enjoyed (without complaint!) the fine vittles and top shelf liquor they poured at every hospitality event at every trade show they ever did. Of course, that doesn't excuse SGI, but sounds a little disingenuous of some of the posters, IMHO.

In closing, I contend that this ending to the story, while terribly sad, is not unique to SGI or even the computer industry. No doubt we will see / hear of such folly again, and probably much sooner than any of us would like. Meanwhile, I fondly remember the friends I had working there, the good times they showed us all (including SW that enabled some of the "older" animated pictures we have all come to love), and the experiences they enabled me (personally) to have, delivering top shelf products to the far reaches of the globe, and being thrilled with how little support they required through life.

SO thanks for the memories, SGI. Not everyone is gonna slag you off now that you're gone.

It is indeed a sad day. See my "obituary" post on Silicon Graphics at: http://subbaiyer.wordpress.com/2009/04/03/sad-about-silicon-graphics/

I did write about the lessons that one can learn from SGI here at: http://subbaiyer.wordpress.com/2007/03/12/lessons-from-sgi/

WANG went BK in 1989. It was a death watch story on the front page of the WSJ. I worked there, had one of the best times of my life and cried when it died in my arms.

Call it what you want - it remains Darwinism without sentimentality. Adapt or die

I was pleased to see E&S mentioned. Shows some have a firm background in the graphics pioneers.

I am sad to see sgi go.

Spent many years at E&S and then a few at sgi. Wonderful times in the 90's, but even while drinking the Kool-Aid I knew the end would eventually come.

Great ride and great technology.

Clark:
I'm so delighted to hear someone hit the nail smack on the head. Being in the field of EE for almost four decades, I worked for successfull companies (good management),on the East coast and the West coast, and have had four companies go "belly-up"under me. In those four instances your words are so fitting that I will reprint them in there entirety.

Clarks words:
Usually what happens at these companies is that management loses touch with reality, focuses on the glory and the money and themselves, and forgets about the employees who were actually the ones supplying the majority of the creativity and energy that made the company what it is. The employees then go into misdirection, many leave, creativity and know how is lost, motivation levels are lowered. At that point, the company starts to unravel and its the beginning of the end. This common chain of events usually never gets observed from outside the company and that story is never told.

AMEN!

Rackable will be doomed if they keep any of the sgi management. Sgi has a long history of keeping the dead wood and dumping the productive. Customer support does not exist. Just try to order a spare part. It takes days to get a call back and weeks to get delivery but only if you are on support. The only customer left is the Gov, and most are running as fast as they can to other platforms. Sad

I sure hate to see all the work in IRIX, the Linux kernel extensions in Propack, CXFS, and so on go down the drain. Sure, some of it has been open sourced, but there is alot of rock solid coding that might disappear and have to be rebuilt. I wonder if Rackable would open source IRIX and all the other codes SGI developed? Then sold commercial support for whatever developed? The single system image kernel tweaks for Linux would be a fantastic contribution by Rackable to open source and the SGI segment could provide commercial support.

I remember thinking NT on SGI's when those boxes came out, yuk! However, the Origin, O2 and Octane systems were rock solid. I wonder who will fill the SSI gap in Linux clusters?

This is indeed sad news. Our company used to sell SGI systems a couple of years back. They will just be renamed but its sad to see such an iconic name disappear.

You can add 3Com to the failed company mix, with that idiot Benhamou's "brilliant" purchase of US Robotics just as dial-up was going away for good! When instead if they'd just improved their weak router line, and maybe not focused so much on ATM with their switch line.

Then there was the Synoptics/Wellfleet disaster. Didn't they know that Wellfleets were only functional as doorstops?? Of course Nortel's purchase took it to a whole 'nother level - Lol! At least the Netgear folks bought themselves out and remain afloat.