Nekochan Net

I found this very interesting and surprisingly detailed, point-by-point rejoinder written by an SGI employee and posted to his own page on the web arguing how SGI was actually doing quite well and was not about to fail.

http://web.archive.org/web/199805111922 ... myths.html

Many of you may have seen this, but it made for a very interesting read. I wish he was right.

_________________
--

That was a great read, thanks!

_________________
33mhz R3k/48mb/XS24 150mhz R4400/256mb/XL24 600mhz R14kA/2gb/V10 8x1.4ghz Itanium 2/8GB 32x600mhz R14kA/24GB 4x700mhz R16k/8GB/V12/DCD/SAS/FC/DM5 2x700mhz R16k/4GB

Thanks for sharing that. I don't remember having seen it before. It was written a few months after I placed an order for my first Origin rack!

We've rehashed the subsequent history of SGI a bunch of times here, but that piece is a very nice snapshot of the way that many people viewed SGI's strengths moments before SGI reached its tipping point into decline. (Of course, one could argue that SGI's fate was sealed when it sold Cray's E10000 Starfire tech to SUN, which predated the post, but I think they still could have beaten SUN if they had the right management vision.) In the end, though, the same fundamental trends that took down the old SGI also took out SUN, so perhaps it's all academic.

I think you can see many of the coming mistakes embedded in that very page. The use of the brand to sell Wintel workstations which were perceived as not extensible or price competitive, then when that didn't work, headlong into commodity PCs with no added value. The transfer of 3D API technologies to Microsoft. Slightly later, in 1998, the sacking of the advanced microprocessor research team, and adoption of the still years until release Itanic. That last move, by 3 out of 5 of Intel's Big-RISC competitors, went a very long way to cementing their monopoly on technical-grade CPUs in this decade.

_________________

The argument made about the cost of an O2 actually being pretty competitive with PCs that came close to delivering the same capabilities is reasonably compelling. They should have stuck with MIPS/IRIX and innovated as best as they could. Would have gone better for them than what actually transpired. They had so many customers in the government space, visualization and elsewhere that the new MIPS/IRIX systems would have definitely found a market, while giving the low-end gaming stuff some time to take root.

Working with Microsoft was a mistake. Opening the kimono and going the OpenGL route was probably a mistake. Adopting the Itanium was another mistake. Just goes to show that open and cooperative models don't always lead to happy endings. They should have defended their unique advantages and innovated on their own platform...

_________________
--

They did, that is why they went out of business.

Just about the only thing these sort of perennial threads manage to accomplish, it is to highlight how people, on the other side of the equation, know very little about how the systems sausage is made, specially its economics.

It's dead Jim...

_________________
"Was it a dream where you see yourself standing in sort of sun-god robes on a
pyramid with thousand naked women screaming and throwing little pickles at you?"

God damn, learn something new every day ! So Windows was SGI's platform and the Itanic* was one of their unique advantages ... I wish I were in the Inner Circle and understood these things the way you do, R10 !

* The same Itanic that HP was paying Intel 400 million a year to keep alive, I just read. Itanic joins Afghanistan in the Graveyard of Empires Association. Way to go Intel ! If you can't actually produce a better processor, you can always fuck with their heads ! American "leadership" is composed of buffoons who get their information from Ziff-Davis anyway.

With all due respect, no they didn't.

They cancelled IRIX and MIPS development. They latched on to a commodity processor. They didn't defend the unique advantages of their own architecture.

_________________
--

You are seeing it from an emotionally vested perspective. Where realities like design and manufacturing costs, profitability, market share, distribution channels, ip, roi, etc. are ignored because what you fetishize is the end product/technology not its process. And that is ok, this is a site about fans of SGI's machines, it is the product that which people here are concerned with after all.

The thing that gets missed is that the same process that created those great products/technology, was flawed. It could not scale to keep up with rising design costs of that which is high performance. That was obvious by the mid 90s to anyone looking at historical design cost data, and doing a simple extrapolation from that curve. The fact that SGI bought MIPS and later CRAY is a clear indication that they, as an organization, were tone deaf and blind to that basic reality.



So SGI did not fail because they started using commodity components, they failed because they held to their proprietary stuff too long. I think it's nearly impossible to compete against a commodity which advances almost exponentially, because the commodity guys eventually catch you and leave you behind far far away in the dust rather quickly. Unless you are riding a similar design cost curve. But when the size of the market is not large enough to finance the next step in the cost curve, then it's over... unless you subsidize from other revenue streams. SGI lacked that.

_________________
"Was it a dream where you see yourself standing in sort of sun-god robes on a
pyramid with thousand naked women screaming and throwing little pickles at you?"

They *tried* to latch on to a commodity processor, and failed.

_________________
(Maradona) (DavidVilla) A1186 (Xavi)
A1370 (Messi) dp43tf (Puyol) A1387 (Abidal) A1408 (Guardiola)

"InfiniteReality Graphics - Power Through Complexity"

What a bunch of Business School nonsense. The reason SGI failed was simple : management had the intellectual acumen of a group of pine bark beetles. If there was any way on earth to fuck something up, they did it. Venture-capital-driven morons who were lucky they had chauffeurs because if they had to find their own cars in the parking lot they never would have made it home at night. And when they finally did accidentally get a guy with an IQ over three, after he turned the company around a little they dumped him.

Typical "hi-tech" Silicon Valley behavior. The dickwads running those companies are good at one thing and one thing only - fleecing the "shareholders" for every penny they've got. Shysters. Ignorant buffoons. Dickcheese imbeciles, that's what you've got running Silicon Valley and now the country. Whoop-dee-doo.

I see what you're saying, but SGI would have had a defensible position and a large enough market if they did not divest themselves of Alias|Wavefront and kept Maya running on their own systems, continued to focus on their throughput strengths (it took YEARS for PCs to catch up to the bandwidth of the Octane - and that was without any development happening on the SGI side of the fence), did not "open" IrisGL and start helping Microsoft on their path to Direct3D/DirectX, didn't sign on to the Itanium and kept developing MIPS and did not execute their numerous disastrous about-turns on whether or not they were serious about the high-end/Cray market.

Net-net, SGI is almost profitable today. The gaps over the last few quarters can be made up with a little bit of belt tightening and by trading slightly slower growth for profitability today. Their lack of profitability (minor, undoubtedly) is more a question of choice right now. They're in decent shape, and the reason for this is that the bulk of their revenue and growth is on the back of SSI clusters. All the software that runs on these boxes got developed for Linux. From scratch. Usually, the software running on such systems is not COTS, it is custom developed. As long as its Un*x, the development effort/cost is roughly the same, not to mention that IRIX had 10+ years on Linux in terms of reliability and stability. It also had a more advanced TCP/IP stack and a better user interface than the competition at the time SGI decided to abandon it. No reason SGI couldn't have delivered the exact same SSI solution on IRIX. It's not like customers needed to run MS Word or had an OS-driven reason to look the other way. As mentioned earlier, SSI clusters run mostly custom stuff. Today, SGI delivers the largest SSI Windows and Linux systems. What if IRIX/MIPS were the only platforms that delivered this level of scale? With IRIX' level of maturity, they could have arrived at the same point faster and would have been able to claim a major differentiator for the platform and justify higher margins.

The management at SGI took decisions based on buzz words like "Open", rather than realizing that their strengths were always a proprietary, highly custom, highly optimized computing platform for a couple of very specific things. There was no need to give away the crown jewels in the hope of gaining low-margin volume. As Larry Ellison has demonstrated, custom and proprietary can be a very good thing. So can suing other people for infringing IP, rather than willingly opening the kimono. Oracle probably justified their purchase of Sun, at-least 50%, because they saw an opportunity to enforce IP rights against Google. And on the other hand, SGI willingly gave everything they had away to their competition. Do a US PTO search and see how many patents were issued to SGI.

Open, "free", "collaboration", "standards"; this crap can sometimes be over-rated and just because people like to write about it a lot and give it a democratic, progressive tinge it doesn't make it good business strategy in every situation.

_________________
--

I think this is the best explanation of SGIs failure. And it's not only SGI, you can see it everywhere from food through automobiles to clothes.

Here's John Carmack's opinion about workstations from 1999:
http://www.team5150.com/~andrew/carmack ... 124x092521
Especially this:

Basically, this is the same what R-tek-K said.

_________________

These threads are the hallmark of all that's pointless about ... Not going to say it.

_________________


DECUS Member 368596

They are pointless in the sense that SGI is gone and will never return, yes.

But if a thread can show people that CEO's are imbeciles, that the Business School whitewash "this is the inevitable trajectory of any business" is ignorant foolish horse crap, if people can learn to use their own minds rather than listening to the Down's children who pass themselves off as "authorities" then they are worthwhile.

There was no inevitability about the downfall of either Sun or Silicon Graphics. Both companies were destroyed by the drooling conventional-wisdom retards running them, just like Bendix, Marrtin Marietta, TWA, almost-IBM, Yahoo, all the places the Romneys and Icahns and Kerkorians ran into the ground so they could walk away with millions .... the United States has become totally perverted, a vicious pit viper eating its own young. None of it is "inevitable", unless people make it be inevitable.

If a thread can convince even one person that the dickcheese imbeciles like Carol Bartz, Carly Fiorina, Bozo Bob Ewald, Ed McMuffin, Mark Hurd, Meg Whitman, John Akers, Tom Perkins, Timothy Geithner, Jamie Dimon, Lloyd Blankfein, Kirk Kerkorian, Bill Agee, etc etc etc are not worth spit (much less 16 million dollars when you finally fire them for incompetence, e.g. Ms Bartz); that they are the cause of the troubles the entire world currently faces ... well, in that case the thread is not pointless.