1. The Cray aguisition
The whole company focused on the highend and integrating Cray's technology and took it's eye off the low end commodity market. 2. Windows NT 3.5.1
The first version of NT that was stable and made the developers of scientific and CAD apps and universities (SGI's bread and butter) to take Microsoft seriouslly. 3. Ego and denial
Tom Jermoluk and Ed McCracken refused to admit there was a 500lb pink elephant in the middle of the board room with Intel and Microsoft logos tatooed on it's ass. Jim Clark didn't, but he lost a power struggle with McCracken and went AWOL until he resurfaced with Netscape. By the time SGI decided to build Wintel boxes, it was too late. 4. The Internet boom
SGI's heyday pre-dated the 'net boom and when the boom hit their talent pool was drained by startups such as nVidia. 5. Re-inventing the wheel
Few of SGI's products ever "evolved". Once they completed rev. 1, they threw it into the market and went completely back to the drawing board to reinvent something brand new instead of evolving (and supporting) existing products. 6. Developing for the sexy, not the practical
SGI's products were sexy and they demoed great but the reality was that the flashy capabilities targeted the small, niche markets like entertainment while they ignored more practical (but un-sexy) features which made them more valuable in larger markets such as CAD. 7. Horribly late in bringing products to market
The O2 and Octane workwstations, as well as their last decent graphics card, were 1.5 - 2 yrs late in coming to market. By the time they were announced they weren't competitive and comprimises had to be made to shoehorn the latest processors from MIPS into them. And the O2's graphics were hardwired to the motherboard and couldn't be upgraded. And as detailed in #6, they were loaded with features that demoed well but almost nobody used (e.g. Octane's crossbar bus technology). And then McCracken handed development of the Wintel box off to the same group that screwed up the O2 and the made the exact same mistakes again. 8. No clue on Wintel
As mentioned above, the group that developed the O2 was handed the keys to develop SGI's first Wintel offering and they made the exact same mistakes. Horribly late to market, loaded with features few people used, and with graphics that were hardwired to the motherboard and couldn't be upgraded. They even went as far as to build a custom BIOS incompatible with existing BIOS-based tools such as Ghost. When sys admins evaluated the boxes, they said "wait a minute.... I can't use my existing admin tools with this thing.... forget it." 9. Fahrenheit
Just as John Carmack had given OpenGL the shot in the arm it needed and it was gaining incredible momentum against Direct3D, SGI made a deal with the devil to merge OpenGL and Direct3D. Fahrenheit never saw the light of day, OpenGL lost it's momentum, SGI's CEO got a high paying gig at Microsoft, and OpenGL is going to be a second class citizen in Vista. SGI also gave up it's efforts on scenegraph and large model APIs which would've differentiated OpenGL even further.
*sigh* so sad
The time I spent working for SGI was one of the most rewarding experiences in my career.
Slashdot Mirror
1. The Cray aguisition
The whole company focused on the highend and integrating Cray's technology and took it's eye off the low end commodity market.
2. Windows NT 3.5.1
The first version of NT that was stable and made the developers of scientific and CAD apps and universities (SGI's bread and butter) to take Microsoft seriouslly.
3. Ego and denial
Tom Jermoluk and Ed McCracken refused to admit there was a 500lb pink elephant in the middle of the board room with Intel and Microsoft logos tatooed on it's ass. Jim Clark didn't, but he lost a power struggle with McCracken and went AWOL until he resurfaced with Netscape. By the time SGI decided to build Wintel boxes, it was too late.
4. The Internet boom
SGI's heyday pre-dated the 'net boom and when the boom hit their talent pool was drained by startups such as nVidia.
5. Re-inventing the wheel
Few of SGI's products ever "evolved". Once they completed rev. 1, they threw it into the market and went completely back to the drawing board to reinvent something brand new instead of evolving (and supporting) existing products.
6. Developing for the sexy, not the practical
SGI's products were sexy and they demoed great but the reality was that the flashy capabilities targeted the small, niche markets like entertainment while they ignored more practical (but un-sexy) features which made them more valuable in larger markets such as CAD.
7. Horribly late in bringing products to market
The O2 and Octane workwstations, as well as their last decent graphics card, were 1.5 - 2 yrs late in coming to market. By the time they were announced they weren't competitive and comprimises had to be made to shoehorn the latest processors from MIPS into them. And the O2's graphics were hardwired to the motherboard and couldn't be upgraded. And as detailed in #6, they were loaded with features that demoed well but almost nobody used (e.g. Octane's crossbar bus technology). And then McCracken handed development of the Wintel box off to the same group that screwed up the O2 and the made the exact same mistakes again.
8. No clue on Wintel
As mentioned above, the group that developed the O2 was handed the keys to develop SGI's first Wintel offering and they made the exact same mistakes. Horribly late to market, loaded with features few people used, and with graphics that were hardwired to the motherboard and couldn't be upgraded. They even went as far as to build a custom BIOS incompatible with existing BIOS-based tools such as Ghost. When sys admins evaluated the boxes, they said "wait a minute.... I can't use my existing admin tools with this thing.... forget it."
9. Fahrenheit
Just as John Carmack had given OpenGL the shot in the arm it needed and it was gaining incredible momentum against Direct3D, SGI made a deal with the devil to merge OpenGL and Direct3D. Fahrenheit never saw the light of day, OpenGL lost it's momentum, SGI's CEO got a high paying gig at Microsoft, and OpenGL is going to be a second class citizen in Vista. SGI also gave up it's efforts on scenegraph and large model APIs which would've differentiated OpenGL even further.
*sigh* so sad
The time I spent working for SGI was one of the most rewarding experiences in my career.