PDA

View Full Version : WIRED: Whatever Happened to SGI?


RobertoOrtiz
11-26-2004, 04:01 PM
QUOTE:

"In California there's a computer manufacturer that makes powerful machines beloved by a tiny niche of creative users, featuring a media-oriented Unix operating system and stunning industrial design. But it's not Apple Computer.

The company is Silicon Graphics Incorporated, or SGI (http://www.sgi.com/), which once was famous for its high-powered graphics and 3-D workstations but has fallen on hard times of late.

SGI now focuses on supercomputers, but there's a tiny coterie of fans dedicated to keeping the company's aging but high-powered workstations alive.
Sites like Nekochan (http://www.nekochan.net/), Silicon Bunny (http://www.siliconbunny.com/) and the SGI Zone (http://www.sgizone.net/) offer hardware and software tips, news, discussion and secondhand trading facilities to the small clique of true believers. Nekochan's bulletin board has 700 members -- impressive for machines were never sold directly to the public. "

>>Link<< (http://www.wired.com/news/technology/0,1282,65834,00.html?tw=wn_tophead_1)

-R

Hazdaz
11-26-2004, 04:03 PM
I read that earlier today... pretty interesting. Unfortunataly I still don't see any use for an SGI machine for doing 3D work (well the ones that would be great cost more than my car). But like the guy said in the article, for video stuff, its probably a good deal.

Andrew W
11-26-2004, 04:13 PM
And they're good for servers too. You can put a quite frightening load on some of these old boxes without them keeling over. I still have a soft spot for my old O2 and I'm toying with buying an Indy for no good reason other than it would be nice to own one.

A

unchikun
11-26-2004, 06:50 PM
I love this quote:

"If the Mac community is dwarfed by the Microsoft horde, the number of SGI users amounts to a rounding error."

*sigh* SGI had so much potential.

I've seen old SGI's converted to all manner of servers, routers, fridges and boomboxes ;) There are definately reliable. I think the reason why Irix has stayed at 6.5x for so many years is because of these old machines. I'm keeping my old Indy thats for sure.

I've talked to researchers and guys who do fluid, weather, seismic similations that love the new workstations for there incredible bandwidth and memory access that beat any pc hands down.... starting at $20,000 they better!

halo
11-26-2004, 07:10 PM
which is bizare because the supercomputers networks that make weather predictions etc that costs millions are made from PCs...

dmaas
11-26-2004, 08:49 PM
SGI's workstation business was killed by gamers. More specifically, their customer base was so small that they couldn't put as much money into R&D and marketing as the mass-market giants like Intel and NVIDIA (and even if they could, the costs would have to be amortized over a much lower sales volume). Intel probably sells 1,000 Pentiums for every MIPS chip SGI sold, and NVIDIA probably sells 1,000 GeForce cards for every RealityEngine SGI sold. High sales volume led to better R&D and the mass-market companies eventually steamrolled SGI from behind.

Apple wants to be the new SGI now (for graphics at least), and I think they are very close to that. They have the advantages of a somewhat larger user base, plus very distinctive software that sways customers who would otherwise buy Windows or Linux boxes.

Recursive
11-26-2004, 10:44 PM
..... its only a tool.
The application is where you do your work.

Use whatever makes your application as good as it can be.

SamChen
11-26-2004, 11:22 PM
:thumbsup: I was fortunate to have been at Sgi even before the web exploded, thru their heydays, and their eventual downfall. They truly had the most innovative products of their time... and for that matter, way ahead of their time. They had clear shell O2 workstations that were available only to internal employess (way before the first Mac went clear).. as well as a chrome plated T2 looking O2 ... which only the Sgi execs had. It was an amazing company to be at.

Speaking of nVidia.. a lot of the core people there are from Sgi... no surprise. It's like they broke away from the short-sighted Sgi execs to do something that was more risky but a worthy one. In retrospect, it was obviously very smart and lucrative. When Sgi bought AliasWavefront... it was amazing to see this synergy.. although short-lived. All of a sudden, you had a workstation out of the box that could do CG films and digital A/V like a dream. This was mid-90's ...way before Macs and Pc's could even come close. So what happened? ...our execs poopoo'ed Windows as a non-threat and was caught offguard compeletely when they came out with NT and cheap graphics cards that rivaled the lower to mid end Sgi workstations. The Indigo, Indigo2, and Indy workstations became non-competitive almost overnight. The beautiful O2 (with its graphics chip hardwire onto its motherboad) was not quite better enough to justify the higher cost. The writing was on the wall.

As a former co-worker puts it, the beginning of the end was when we threw extravagant xmas parties with entertainers like KennyG, Patti Labelle, & Huey Lewis. Even Clinton & Gore (inventor of the internet :scream: ) came by to launch technology initiatives on our campus. Good ol days indeed. Now, I look back fondly and still reminisce about the days when SIGGRAPH was dubbed SGIGRAPH where our booth had the most consistenly exciting things each year. So much change in so little time, as I'm writing this post on my Dell Inspiron 8500 laptop w/ GeForce4 graphics that can run circles around $50,000 Octane workstations from 5 years ago.

dmaas
11-27-2004, 12:43 AM
Cool info, thanks...

I certainly wouldn't send back a Fuel or even an Octane if one showed up on my doorstep :)

SGI's "Digital Media" API is awesome. To this day you cannot buy a non-SGI system that has all its capabilities. It makes Quicktime look like a toy by comparison. (SGI was doing uncompressed HD video since before DV became popular :) ).

I have an old Indy I keep around for fun. Its 2D X server was faster than the Linux X server on a Pentium III.

Another thing SGI really got right was scalability. On a PC you are usually limited to one video card, one or two SCSI cards, one or two network cards, etc. Whereas you could cram as many of those as you wanted into a modular SGI system.

unchikun
11-27-2004, 04:37 AM
The scalability came from SGI's aquisition of the "supercomputer" company Cray. It is quite the feat to share resources across multiple machines.

For a couple of years SGI didn't even have a booth at Siggraph. Goes to show the downward spiral the company went through. I would see some SGI/Alias sales people or middle management at a company bash with nice leather jackets with the company logo on it and think "Ah, this is where my annual maintenance support money goes".

wmendez
11-27-2004, 07:23 AM
I've seen old SGI's converted to all manner of servers, routers, fridges and boomboxes ;)
Like Ed's Cabinet
http://www.edharriss.com/misc/sgi_cabinet.html

SGI Fridge
http://home.planet.nl/~mourits/koelkast/

s_ridenour
11-27-2004, 08:49 AM
...our execs poopoo'ed Windows as a non-threat and was caught offguard compeletely when they came out with NT and cheap graphics cards that rivaled the lower to mid end Sgi workstations. The Indigo, Indigo2, and Indy workstations became non-competitive almost overnight. The beautiful O2 (with its graphics chip hardwire onto its motherboad) was not quite better enough to justify the higher cost. The writing was on the wall.
I thought the problem was that "Chainsaw Rick" decided that the company should let MIPS and IRIX languish and focus instead on making Wintel boxes.

Of course, I didn't work at the company, so you probably know a lot more about it than I do.

Hazdaz
11-27-2004, 02:50 PM
I lost a decent amount of money on SGI - I bought some stock right before they came otu with their NT-boxes. I was (mistakenly) thinking that their NT-boxes would kick ass and help their sales... but little did I know how poorly managed the company was - even if they did have great technology ( I was atleast right about the technology kicking ass).

They were way ahead of their time, and only now is some of their technology going down to "regular" PCs. Its a shame how many computer companies throughout the past 20+ years have been in the same position as SGI - great technology, but poor management/marketting.

tmcbroome
11-27-2004, 04:38 PM
Many of us have witnessed this cycle before... an amazing machine, an amazing OS (though the GUI for IRIX was slooooowwww), an amazing development department and rabid supportors; all killed off by greedy Executives that are simply there for the ride, paycheck and "retirement" package. I am of course talking about the Amiga as well as SGI. SGI had some nice boxes but the fact is that when NT was mixed with PIII's and a decent video card they were non-competitive. I could get 2-3 times as 3D work done over on the damn wintel as could on the sweet Indigo's or even O2's.

I am sad that SGI was not able to move with times but I am happy that workstations became cheap enouph to be a household item, very happy indeed.

SamChen
11-28-2004, 10:48 PM
:wise: What a lot of people forget about Sgi is that they were such innovators of software as well. Because their hardware was so "open" and innovative, the software engineers there pioneered some of the most amazing things to take advantage of the unique hardware. Things we take for granted today like OpenGL, digital media, GUI for Unix, advanced image processing, texture roaming of huge textures, advanced real-time level of detail switching, teleconferencing, video on demand, and the list goes on. Don't forget that they also provided the foundation for VRML (Virtual Reality Modeling Language) which was the standard for 3D on the web. And of course, MAYA was developed at its subsidiary AliasWavefront during this meteoric rise as well as their involvement with the Nintendo64. So much richness in ideas and innovations in such a short burst of time. It was one of the most exciting places to be at in Silicon Valley before the dotcom era.

As for "chainsaw Rick"... it really wasn't just Rick Belluzo's fault. He was brought in the help stop the bleeding, but the internal hemorraging was already started when the previous CEO Ed McCracken and founder Jim Clark didn't see eye to eye about its future direction and the threat and opportunities of the low end. As a result, the bottom dropped out when NT workstations like Intergraph et al. started chipping away at their market share. By then, the innovative Sgi Visual Workstation 320 and 540 were too little, too late. Rick Beluzzo took a lot of the blame and was quickly booted. By then, Sgi was too deep in its own grave to climb out. 20/20 hindsight. Now, it can only attempt to hold on to its dwindling niche.. which is high margin supercomputers and servers.

By the way, here's a great link for what people have done to old Sgi boxes. Lots of the links are broken.. but you get the idea: http://www.bbsolutions.com/jodeman/unusual.html

enjoy!

ages
11-29-2004, 04:00 AM
Were there any reputable 3D benchmarks of SGI machines vs PC's?

I have always longed for the days when macs and pcs were dedicated to one genre...like Amiga and SGI were.

Stahlberg
11-29-2004, 05:51 AM
I doubt if they're actually dying. They've lost a lot of lower end business, but their most lucrative game is still doing quite okay - government contracts. most importantly US military.

And - time for a little rant - let's not forget revenue from selling governments of un-developed or developing nations huge multimillion-dollar "show-off" graphics systems, obsolete of course.
The system is brilliant in its simplicity. Just go to a country that's trying to upgrade it's IT status (and which isn't?), find the highest decision-maker that is susceptible, and offer them a % of the sale. You'll find one eventually, even a small % will be a huge windfall to such a person. It works like a charm, because SGI has the past reputation and the hardware in the glossy catalogs to put substance behind their claims, and the decisionmakers are all laymen who know next to nothing about computers. Sometimes I guess they might search for another company to offer a competing bid (although I imagine they'd hesitate to loose that nice kickback), but there's no real competitors when it comes to supercomputers so... :shrug: taxpayers money squandered, suits get richer, IT program set back. Some international agency should investigate. End of rant.

fxgogo
11-29-2004, 09:53 AM
I remember seeing that travelling truck they had (this was in South Africa), and they had some realtime 3D demo showing a picture and you could select any part of the screen and strech it like rubber. I was blown away. We live in a funny industry where past achievements seem so paltry now, but we forget all too quickly how slow and less complex it used to be.

Als
11-29-2004, 04:29 PM
Well, I guess I wonder why nobody mentions the fact that main problem
was the processors speed. What's the reason for this is mistery to me.
Why MIPS could not go faster is main reason.
Ok the GUI could have done better, but there were many good points
SGI machines had.


Al

EdHarriss
11-29-2004, 08:44 PM
Like Ed's Cabinet
http://www.edharriss.com/misc/sgi_cabinet.html
- Ha! Will beat me to posting my own Sgi. :)
- BTW, I've got an old Indigo Elan that I'm planning on turning into a mini stereo or DVD player.

beaker
11-29-2004, 08:57 PM
What people in cgi forget is that our business was only 5% of SGI's overall market. They wasted huge amounts of money and resources supporting the cgi industry when we only made up a very small percentage of their overall sales. They mainly did it for the press and fame, not for money, which wasn't a very smart business plan.

MasterBaiter
11-29-2004, 09:01 PM
I have 4 o2's collecting dust in my garage right now. :sad:

Kabab
11-29-2004, 10:50 PM
Its funny every design/engineering firm i visit has O2's and octane's lying around doing nothing but no one can bring themselves to throw them out :P

Andrew W
11-30-2004, 09:04 AM
Its funny every design/engineering firm i visit has O2's and octane's lying around doing nothing but no one can bring themselves to throw them out :P Ah yes the ever popular 15 000 Octane footstool or door stop. As seen in studios the world over.

"Pull up an Octane and sit down".

A

PS I still own and won't give up my o2, and I'm toying with buying an Indy. That was the first 3D workstation I used way back when...

SheepFactory
11-30-2004, 09:12 AM
i had a Sgi 320 and not so fond memories with it.

the Sgi widescreen lcd was a beauty though.

benvu
11-30-2004, 11:58 AM
My first animation was completed and output to video (uncompressed) on an SGI 320 running Premiere something. I remember I had to use a QT output written by some guy to get it to output correctly. This was before I heard of something called a 'firewire.' I still hate to watch DV compressed stuff. Does not do justice to our hardwork.

Still craving that SGI 540 with Quad CPUs! Still hard to find those puppies on ebay.

Thalaxis
11-30-2004, 02:55 PM
I doubt if they're actually dying. They've lost a lot of lower end business, but their most lucrative game is still doing quite okay - government contracts. most importantly US military.

The news is actually better for them than that, because their government sales have been dwarfed by
their commercial sales. Quite a few companies have been deploying SGI Altix machines with the likes
of Oracle and SAP for enterprise resource management and that sort of thing.

Another bit of good news is that one of the SGI guys on a tech forum I frequent has been hinting at an
upcoming workstation product from SGI... it's not definite (he can't give out details yet), but it sounded like
it was going to be a "small" version of their new Prism visualization system.

Als
11-30-2004, 03:31 PM
There is Indy on ebay.co.uk for 16, Indigo2 is 100.
Nice! :)

I guess all SGI needs to come up with is strong processor
from MIPS which would run on 3Gb, and they will be back
in the game again...


Al

Ah yes the ever popular 15 000 Octane footstool or door stop. As seen in studios the world over.

"Pull up an Octane and sit down".

A

PS I still own and won't give up my o2, and I'm toying with buying an Indy. That was the first 3D workstation I used way back when...

Thalaxis
11-30-2004, 03:48 PM
I guess all SGI needs to come up with is strong processor
from MIPS which would run on 3Gb, and they will be back
in the game again...

SGI sold off MIPS a long time ago. MIPS is making quite a bit of money with their designs in embedded
and networking products as well as in licensing. They haven't been trying to develop a high-end
processor for the past few years, so they'd be far to late to the party to have a fighting chance at it now.

SGI's making a slow, steady comeback, and they're planning to release some new gear based on
Montecito and Silvervale next year to keep the trend going. If they don't drop the ball, it could work out
quite well. And if they do launch a solid workstation based on Montecito or Silvervale and price it
competitively, it could put them back in the game decisively -- especially since they're largely a Linux
shop now, and they have already developed a solid graphics infrastructure which they launched with
Prism.

maninflash
11-30-2004, 04:13 PM
Couldn't SGI just make their own low-mid end workstation with reasonable price accessable to amaturs and pros and start gaining back the lost ground? Nothing fancy, just something that could rival a dual xeon system with Nvidia graphics? I think if the price and performance is about the same, most people would choose SGI becasue of its history in computer graphics, ...I definitly would!

Als
11-30-2004, 04:17 PM
Thanks!
I was always puzzled by the fact that MIPS was stoped in development of processors, I didn't catch the moment when it was sold by SGI.
Once on the computer fair, I spoke to one of the SGI guys and he said
that they were planning comeback with 1200 MHz, (in time of 2Gb intels if I remember corectly), and then nothing really.
Any insights on that?
I like SGI operating system, not so much GUI, but there are alternatives. Also main hardrives were quite small size, and quite
expensive, which was really stupid move as far as I can see.
Still the great thing was that after machine crashes badly (which I've seen five times in 15 years) you just reinstall operating system and you are open and running.


Al

Thalaxis
11-30-2004, 04:41 PM
Thanks!
I was always puzzled by the fact that MIPS was stoped in development of processors, I didn't catch the moment when it was sold by SGI.

MIPS didn't stop... they kept going after SGI spun them off, but not in directions thta would be of any
use to SGI now.


Once on the computer fair, I spoke to one of the SGI guys and he said
that they were planning comeback with 1200 MHz, (in time of 2Gb intels if I remember corectly), and then nothing really.
Any insights on that?

They took stock, looked at what Intel/HP were doing, and decided to let Intel shoulder the financial
burden of developing ever faster processors. At that point, they (ignoring HP's example), cancelled
their high-end MIPS projects, sold off MIPS, and decided to transition everything to Itanium.

History shows us that the Itanium has in fact delivered on most of its promises... but it did so close to
six years later than SGI and Intel and planned on, and unlike HP, SGI didn't keep their in-house
designs on at least extended life support to give them something to tide themselves over until they
could realistically make the transition. HP was much smarter; they announced their plan to transition
their products from PA-RISC to Itanium, and phase out the PA-RISC, but they kept active development
going at least enough to show some forward progress on the product while the Itanium came online.

Now, combine that with the change in the industry, and it's just another nail in the coffin for the SGI of
old, because that blunder was in addition to a dismal failure to adapt to the times; back then,
supercomputing was a HUGE chunk of the high-end market's dollars. Now, while supercomputing still
involves a lot of money ($9billion last I heard), it's quite small compared to the enterprise market.


I like SGI operating system, not so much GUI, but there are alternatives.

Altix runs Linux kernel 2.4 now, with 2.6 in prep.


Also main hardrives were quite small size, and quite
expensive, which was really stupid move as far as I can see.

I was always rather amused by their pricing, especially when I took the model numbers off of one of
their drives (a Seagate Barracuda) and priced it at DCDrives.com.


Still the great thing was that after machine crashes badly (which I've seen five times in 15 years) you just reinstall operating system and you are open and running.

Let's hope that they've improved on that. One thing SGI's never been good at in the past was building
enterprise-worthy systems. And when your system manages millions (or in the case of companies like
Fanny Mae and Freddie Mac billions) of dollars a day, downtime is BAD. :)

Thalaxis
11-30-2004, 04:49 PM
Couldn't SGI just make their own low-mid end workstation with reasonable price accessable to amaturs and pros and start gaining back the lost ground? Nothing fancy, just something that could rival a dual xeon system with Nvidia graphics? I think if the price and performance is about the same, most people would choose SGI becasue of its history in computer graphics, ...I definitly would!
I imagine that they could, if they took the time to do it. It's not like they don't have the engineering
prowess to pull it off or something, so the only thing that might prevent them from doing that is resources.

What I would like to see is a Prism product built with two processors using Intel's $850 Itanium2 product
(it's basically identical to the flagship model but with 1/3 the on-die L3 cache). That could put it in Xeon
territory as far as price, but Xeon couldn't touch its computing power.

The only real issue for that sort of product would be software. The price/performance ratio is
irrelevant if it doesn't run the software people want to use. At least it's a Linux system, which gives it a
bit of a leg up in that regard, but there's still a serious chicken/egg issue to deal with there, and that's
a lot of resources to expend just for the "if you build it, they will come" hope. And Intel's publicly
backing off from the lower echelons of the market with Itanium, so that might end up nixing the $850
Itanium, in which case this sort of system would end up starting at around $10k, for a barebones
configuration, and I get the impression that SGI doesn't have the money to do parallel x86 and
Itanium development and QA.

dmaas
12-01-2004, 12:02 AM
If SGI wanted to build a low-cost workstation it would have to use Xeons or Opterons (or possibly the G5 PowerPC chip) to be competitive on price/performance. But then they'd be selling essentially the same thing you can get from Dell or Apple, and SGI can't compete with them on price alone due to the disparity in sales volume.

That's what killed SGI's NT workstations - I'm sure they were great products, but they weren't enough better than a no-name Intel box to justify the price premiums that SGI wants to charge. It is very difficult to justify paying $7,000+ on a workstation when Dell and Apple sell their top-of-the-line models below $5,000. UNLESS it includes some kind of special proprietary software or support and maintenance.

I think if SGI really wanted to sell workstations, they should develop a custom distribution of Linux, based on the open-source foundation but also integrating all the extra tools you need for CG work (e.g. a flipbook player, a good paint program, some kind of video I/O standard). Many of these things exist in IRIX already and could easily be adapted to Linux. This would fill the void of a standardized Linux platform for graphics work - which has been missing ever since RedHat dropped its free workstation distribution. (and RedHat never included enough graphics tools out-of-the-box). This would help big studios by creating a standard OS that software vendors like Alias and Pixar can support, and it would help smaller studios by enabling them to take advantage of Linux without having to customize a generic Linux distribution with CG tools.

Apple is sort of taking this approach now, but in their own way.

Als
12-01-2004, 01:32 AM
I asked that same SGI guy about it (read previus),
why they don't make their own version of linux,
which would be really close to IRIX, but he couldn't really
answer. Maybe there are legal issues or they were afraid to scare people of, or who knows.
After working on that misterious stuff with Microsoft, called I think Farenhait, or something, maybe they got burned, or all people went off to work on windows... so they don't have those same people who could do it anymore (or they maybe work on red hat, or suse... )
I think your idea is great, I hope someone there thought about it, but who knows what is stoping them. Maybe it's same uncreative management who is holding back...
I'd like really to see SGI comeback, it would help competition too.
In the glorius times of SGI all other companies hated them, and wanted them to go down...

Thalaxis
12-01-2004, 02:08 AM
To some extent, they DO have their own version of Linux. If they launch an x86 or PPC
workstation product with a CG oriented feature set, they'd be going up against Boxx and
RackSaver and also IBM, though IBM's distro is more geared toward enterprise than CG, in
spite of their being a provider of x86 hardware to WETA Digital (running Linux, of course).

And WETA's also using SGI's InfiniteStorage systems, so SGI isn't starting entirely from
scratch with Prism, at least :)

As far as their Linux distro, they've added a lot of stuff from IRIX to achieve high scalability;
they have the only Linux distro out there that can run on a 1024 processor shared memory
system, and it performs extremely well, based on the feedback from NASA's engineers.

Hazdaz
12-01-2004, 02:41 AM
If SGI wanted to build a low-cost workstation it would have to use Xeons or Opterons (or possibly the G5 PowerPC chip) to be competitive on price/performance.
...UNLESS it includes some kind of special proprietary software or support and maintenance.
.
I think that sums it up pretty well right there..
The problem (and yet also the advantage) of the WinTel platform is that pretty much everything is standardized so its compatible.

Like a new fangled technology can't just be plopped into a system and be proprietary to only SGI or any other company, or else no one is going to support it - thus making that technology pretty much useless. Like (hypotheticallty) if SGI came out with the coolest, fastest, most powerful video card (and didn't cost a ton of $), but it was completely incompatible with D3D or even SGI's own OpenGL, then seriously what good would that card be?

On a (semi-)proprietary OS/hardwre like its own line of *NIX workstations, SGI can just write its own code and integrate it into its own systems. On the WinTel side of things ya really can't do that. (The whole transition to 64-bits is a good example... without MS's blessing, AMD can't show off the true power of its new chips, and Intel is basically being 'forced' to follow AMD's lead since MS wasn't gonna write 2 version of 64-bit Windows)

I think because the WinTel platform is so "generic" and non-proprietary, it really gives companies like SGI a tough time since they really can't have their hwardware be THAT much faster than anyone else - cuz it really can't be that much faster or else its not gonna be compatible with anythying.

Thalaxis
12-01-2004, 02:52 PM
I think because the WinTel platform is so "generic" and non-proprietary, it really gives companies like SGI a tough time since they really can't have their hwardware be THAT much faster than anyone else - cuz it really can't be that much faster or else its not gonna be compatible with anythying.
That was one of the problems with their NT workstations. They were rather unique in that
they used a UMA design with boatloads of bandwidth, but they also had things like the
graphics chip hard-wired. When it hit the market over a year late, it was already outclassed,
so it didn't really have a leg to stand on.

CGTalk Moderation
01-20-2006, 12:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.