Slashdot Mirror
SGI Arises From the Ashes
eldavojohn writes "Six months ago, Slashdot reported on SGI's filing of Chapter Eleven Bankruptcy. I wondered why Slashdot kept the Silicon Graphics category with them now defunct. But Chapter Eleven means a reorganization — not liquidation. And, surprisingly, SGI has dusted itself off and stood back up. What did they dust off? About $150 million worth of spending a year. Will this reorganization put them back as a player in the graphics game? Maybe but as the article notes, they have some stiff competition that offer comparable services for less money. Is this a phoenix story or the final death throes of the company?" To be honest, no one here suspected a thing. We just keep the old topics around so it's still possible to find old stories related to them. Sometimes (like now!) they even still come in handy.
I don't think it's simply a matter of making the hardware, but having the brains left to design it. SGI once came out with the greatest stuff, but now loads of that all fits on one video card or multiple video cards with shared GPUs. Of course their old business model wasn't just to sell you the machine, but to license the software, operating system, sell support etc. Not many can do that these days, like they did in the days of yore.
We just keep the old topics around so it's still possible to find old stories related to them. Sometimes (like now!) they even still come in handy.
Call me a dreamer, but I keep hoping some day these guys will arise from the ashes of HP/Compaq and Intel.
Introducing the PDP-11/128 and the VAX 9990! (2-AAA cell batteries not included.)
A feeling of having made the same mistake before: Deja Foobar
...then I hope to god that they put Irix in line with the OS capabilities of this day. I have to support a small fleet of SGI Octanes running Irix 6.5, and damned if those aren't the slowest and most aggravating machines.
Many of the Alpha engineers transitioned to AMD. That's why we've seen such great developments from AMD over the past few years. While Intel was fucking around with the failure that became the Itanium, AMD had some of the greatest processor designers ever working on the Opteron. And the end result is as would be expected: the Opteron is the premiere general purpose processor around.
And so are the MIPS family of processors. So are many of SGI's core businesses, like selling to the TV networks (now it's Apple-to-Avid with new stuff that simply buries SGI), stringing clientele along to the tune of numerous significant digits for incomplete and ill-designed systems.
The fact that they couldn't hold onto employees because their situation was untennable, with so many chiefs and so few worker bees, may now be changed. It's unlikely that their re-emergence from CH11 will do much to save them. Their emporer still has no clothes and is still charging by the pay-per-view model.
---- Teach Peace. It's Cheaper Than War.
I think it would be great to have another player of graphics. As it is, we only have NVidia and ATI (who both make quality products). However, if there was another player that could do more general-purpose cards (as opposed to gaming), they could probably make some decent money, and indirectly pressure NV and ATI. Maybe open-source drivers? How about a more general-purpose parallel floating-point unit that could be more utilized? Something like the F@H GPU client, but for more apps? As a side effect, it could do graphics.
This, however, is probably wishful thinking. Oh well...
I have developed a truly marvelous proof of this comment, which this signature is too narrow to contain.
Well, they do have new management now. As far as the smart people...
they still know their stuff.
Think of them as a new company, that sells new/different products.
So, speaking as a former SGI Employee (and stockholder). Are my worthless shares still worthless? I used all my shares as tax-write off years ago and prompt forgot about them.
They did a great job pissing away my 5000 share stake at $25 a share. I was writing that off for five full years and the stock is still worthless. I think from what my accountant said their old shares are offically not worth anything and are just empty bits on a brokerage account somewhere.
Thanks for the fuckover, sgi.
Namely, concentrate on the open-source market. Contract with NEC to build a cheap ARM processor on a really old technology using 0.8 micron. Then, build a nice computer around the ARM processor. Use electronic parts that are based on old technology. All the ICs should be 0.8 micron or larger. You can get 0.8-micron chips from China for dirt-cheap prices.
You can probably build a computer for $100 or less. Tune the computer for Linux. Rename the company to International Mozilla Hardware (IMH). Establish the company as a non-profit corporation, not non-profit charity.
There is a huge market for 2nd-tier computers. That market consists of the American underclasses and most of the 3rd world: Mexico, etc. The last laugh will be on people who paid big bucks for 1st-tier computers: the power of a 2nd-tier computer is more than adequate for the #1 Internet application: e-mail.
We met with the SGI salesmen the other week. They confirmed the following:
1) IRIX is dead (SuSE will be used instead)
2) MIPS is dead (high end chips are itanium)
3) SGI graphics products are dead (go buy ATI)
If you're an idiot or a government contracter, they will still specially-engineer such systems for an obscene amount of money (technically, none of these are dead if you are the government with a service contract).
The new SGI will be selling fancy Itanium systems on the high end and basic Woodcrests linux clusters on the low end.
SGI still has extensive experience and knowledge building high-processor count boxes that act as a single system image. They're one of the only players who will sell you an entire rack of nice Itanium systems - oodles of processors, RAM, and ultra-large bandwidth - packaged nicely. If a multi-threaded application requiring > 100 GB of RAM is your bread and butter, they're still here for you. They also will integrate FPGAs directly on the same interconnect as your processor - not even IBM is doing that for general customers yet.
If they are to survive, it's working with these fancy uber-fast, uber-bandwidth interconnects between processors that allow large NUMA computers and having first-mover advantage with Itaniums and FPGAs on a none PCIe/PCI-X bus.
The only software they will be doing is anything directly related to getting these goals accomplished. No more compilers, debuggers, graphics software, OS, or (probably not) file systems for them. XFS will be maintained (and added to by the community, of course), but don't expect SGI-funded XFS2 to appear any time soon.
Overall, they've done a damn good job of cutting the fat and coming up with a roadplan for the future. The only downside is the fact they've put so much money into the Itanium that the company would sink if Intel cut the cord.
SGI still has some really slick tech like NUMA. I work in high-end visualization, and I can tell you that it's tough to get good computers for it. I'd love to see them start combining commodity and proprietary to make new-wave supercomputer hybrids for visualization. The problem is, that like all Unix makers, they think it's better to do everything themselves and lock-in their customers rather than competing. Get over this and they might have something.
People who think they know everything really piss off those of us that actually do.
Such as the lead chip designer for the Alpha, Dan Dobberpuhl. A few others are also listed at http://www.pasemi.com/about/team.html
The PWRficient family of PPC processors is actually very interesting from a HPC standpoint; it may even be of some use to SGI. These chips are fast, extremely low power, and have a ton of integrated I/O and memory bandwidth. They are the perfect chip for an extremely high density Blue Gene style system. (Among many other things.)
In any case, the demise of the Alpha was truly a shame. As for SGI, I believe that their fate was sealed when they changed their name and logo. To discard such a logo is unforgivable; if they were to restore it though, perhaps they may rise again...
Go and have a look here if you haven't already. There's some great stuff.
XFS is an awesome filesystem, and has been ranked the overall best in at least two fs benchmarks:- here, and here. Given what I've read here, I'm possibly considering making it my own default fs...at least for some things.
There's also some OpenGL related projects, as well as some kernel work. What this could also mean for them is that even if they do have to sell SUSE clusters, they can still have some individuality in the offering. Sure, anyone can burn xfsprogs to a CD...but SGI can still market themselves as the people who invented the fs, and thus the people who are most intimate with the code, and thus who can possibly most quickly/easily extend it, or fix it if something breaks.
Uh, yer history is wrong. all wrong. DEC never got the cash. Intel contacted their "buddy" Compaq to buyout DEC, and shutdown the lawsuit. Poof goes DEC and everything else, and all of Intels troubles soon vanish.
If I was going to post a lot of rubbish like that, I'd do it under AC, also. I'm assuming this is actually a troll, but I'll bite anyway. The suit concluded, out of court long before Compaq entered the scene. There was no judgement to go poof and Compaq would be absolute fools to let, IIRC 425 million $ go away, not and keep their executives anyway, no board would buy a company and forgive a large settlement like that.
Intel went on to manufacturer later Alphas under their agreement with DEC before they closed down the fab.
A feeling of having made the same mistake before: Deja Foobar
Numalink was called CrayLink on some Origin2000 systems. But the Origin and its NUMALink was designed well before the Cray acquisition took place. It was just a marketing ploy when they put the revered Cray name on the linkage for the Origin systems.
A 64 node Origin2000 was delivered to my site at Cray right after we were acquired by SGI. The name CrayLink was printed in large letters on the horizontal bar in the middle of the O2K rack.
I thought it was cute, but odd.
Few days ago I saw a piece of news in futuretech.blinkenlights.nl which showed loads of Tezros, Fuels, Octanes and other system on their way to be recycled. Brand-new gear, still in their boxes on their way to be destroyed. Maybe they are not competetive anymore in the price-performance arena, but they are still very very cool machines. Was their destruction really the best option here?
Lesbian Nazi Hookers Abducted by UFOs and Forced Into Weight Loss Programs - -all next week on Town Talk.
What I want is for them to design hardware. I don't care about Irix, I don't care about MIPs. I care about how beautifully systems like the O2 came apart and could be put together again. I've yet to see any PC, Apple included, that was so well put together. We need to get past clunky systems with tangles of wires everywhere to get to properly integrated components.
The problem with Prism isn't that it's expensive.
The problem is that they're extremely slow where your app isn't multithreaded (some algorithms just aren't multithreadable) as the Itanium is just dog slow. Also, despite the high memory, the limitation of inferior (at the time) ATI graphics cards on a non-PCI-express bus didn't help either. So you could have a massive ammount of data in memory, but once you tried to display it, it was slower than on a Sun/IBM/HP opteron box. Lastly, one single byte of data failing within the box causes very fun/hard to debug problem and is not uncommon.
At least the Prisms run a modified version of Redhat (and not Irix). And yes, they've let everyone in the Prism group go except some people for support as far as I'm aware. Long live SGI, the uh...disk storage solution.
(Posted as AC for my own protection;))
-- Political fascism requires a Fuhrer.
After them dropping my stocks with no compensation, why in the hell would I be willing to buy their stocks again? As far as I am concerned, I was a loyal supporter and they royally screwed me over. Screw SGI.
http://www.google.com/profiles/malachid