Slashdot Mirror
Feature: Myth of the Fall of SGI, Part II - the Mystery of Irix
Myth of the Fall of SGI, Part Two -
the Mystery of IRIX
When I last addressed this topic, more than a year ago, I was hoping to clarify and settle the notion of SGI's supposedly then-imminent demise. I did hope it would work.
However, there is still a peculiar occlusion perceived in SGI's future, particularly by those that do not understand what SGI does; and now, more recently, with SGI's logo change and Linux embrace, the former faithful question SGIs existence as well, possibly because of the recent layoffs (despite a strong quarter for profitability), the corporate image makeover, the MIPS stock divestiture, and the spinning off of Cray and their NT division into separate (but still quite owned) companies.
That these perceptions are still unfounded, and have been unfounded for more than a year, that SGI has posted profit recently, and that SGI continues to be a strong seller of UNIX systems despite the disappointing performance of the NT machines. Some people just can't let go of their own bad ideas. It's like the flat earthers, who continue to insist the earth is not round: I expect to run into people who will continue to insist SGI is going out of business any day now for the next ten years and beyond.
I have, at last, accepted this: that I cannot change everyone's mind with reason or facts. But you can get some of them here.
Common Mistakes and Misinformation That doesn't mean I have to let the FUD slide. Therefore, I have a list of common errors and mistakes people make when discussing SGI.
1. SGI tried to convert customers to Windows NT (or Linux).
This is obviously false, as even the most casual visit to their website shows. OSes supported by SGI are UNICOS (Cray), IRIX (MIPS), Linux (IA32/64), and Windows NT (Visual Workstation).
I find it interesting that anytime SGI adds an OS to its roster of supported platforms (UNICOS, IRIX, Linux, and NT), people assume everything will be rolled over into that one at an unspecified date. This is simply not going to happen. IRIX will remain on their MIPS-based desktop workstations, and it will continue to be upgraded as the best technical high-end UNIX out there.
2. SGI has a gloomy future. (SGI is dying, losing all its customers &c.)
This one's my favorite, simply because it can't be proved or disproved. There's no best response to this one, because it's primarily FUD, since asking for proof is generally ignored.
3. IRIX is being abandoned.
Again, another falsehood. SGI never claimed this, and SGI's IRIX support has actually remained unchanged. The only possible difference is looking like IRIX might not make it to Merced after all. Big deal. I have little faith that porting Solaris to Merced, for example, will do much to Solaris's overall market share.
3a. IRIX is unstable/insecure/unreliable.
I don't think these sorts of comments are relevant. ANY UNIX can be unstable, insecure, and unreliable, so why should IRIX be any different? Competently administered, IRIX is an extremely stable, secure, and reliable UNIX. People seem to forget we still do need people behind the console to keep things working smoothly, and a great deal depends on how well that person knows his or her system and how good she or he is with it.
Like any UNIX, IRIX isn't idiot-proof. Nor is it meant to be; most UNIX operating systems aren't meant to be run by even incredibly smart NASA monkeys, but by competent professionals.
4. Adopting Linux is the sign of a dying, desperate company.
Really? Red Hat must beg to differ. As must, to some extent, IBM, Caldera, Penguin Computing, VA Linux Systems, and indeed any business that has adopted Linux in one form or another as a part of its operating model.
5. SGI's graphics performance isn't cost effective compared with gaming cards for PCs.
I have to raise this issue. Tim Sweeny of Unreal (www.unreal.com) made comments in early July about the future of SGI, and how it doesn't stand a chance against gaming PCs. This is absolutely laughable, and Sweeny ought to have known better than make such statements.
Quoth he on his website: Now, a $2000 Pentium III PC with a Voodoo3 or TNT2 card eclipses the performance of a $30,000 SGI for real-time rendering. The CPU is faster, the fill rate is faster. When I first saw an SGI Reality Engine, my impression was, "Holy cow, I can't believe how much better this is than my '286!" But nowadays, the best 3D games look far cooler than anything you see running on a Reality Engine. When I first read that paragraph, I wondered how a veteran of the game authoring industry could make such a carelessly crafted statement without thinking about it. First, nobody buys an SGI Crimson or Onyx for gaming. Never have, never will. Second, nobody makes games for the SGI--people create games *on* the SGI, not for it, because there would be no point in doing so. Third, I'm nonplussed by the notion of any company in the computer industry has to make as many workstations as Dell makes PCs to stay in business, when the last ten years have shown that you can, and most do.
PCs that have OpenGL cards used primarily for gaming are good a small numbers of polygons, and fast pixel-fill rates. Even on the PC, cards good at large-poly-count geometry transformations tend to suck at pixel-fill (2d) operations (check out any decent card for the PC manufactured by Evans and Sutherland if you dispute this). Why this is even confused with SGI's machines is beyond me, since no PC has even been able to match the aging Reality Engine gfx first made available on the original Onyx in 1994, much less the 88 million polys/second (shaded, textured, and antialiased) of the Infinite Reality on the Onyx, introduced two years later.
But, to the lower end. If you're playing games, you're not buying an SGI. If you're creating them, you're probably not buying an SGI, unless you want to create something along the lines of Myst or Riven, the latter of which was generated on SGI machines running Softimage. But, I have noticed SGI usage popping up in more realtime-oriented titles like Spyro the Dragon (my favorite; from Insomniac Games, check 'em out) for modeling.
Most consumer gaming cards are better than all but the highest-end SGI systems for pixel fill rate. But that's all. The instant anyone needs to transform, rotate, or scale more than a few thousand polygons per second while working with unoptimised geometry (for example, during modeling), any modern SGI will beat any consumer card, period.
Just looking at the behind-the-scenes creation of any major motion picture that involved special effects to any degree reveals that SGI is still the first fallback. Why? Probably because that's where the workstation grew up: in high-demand production environments like ILM, Boss Film Studios, Pixar, and Disney (to say nothing of NASA and other supercomputing institutions nearly everywhere). 6. SGI's new logo is just another example of how SGI is dying.
This one is purely subjective. It's based purely on the espouser's opinion, which I contend is largely useless as a legitimate means of examining a corporation's health.. SGI changing its name to SGI was probably a good move, and I like the new logo only somewhat less than the old one.
To be honest, though, I don't care what the silly thing looks like. I don't see the logo as having much impact on the speed of the systems themselves, the stability of the OS, or the price of tea in China.
This sort of comment reminds me of the people who used to post flames to Usenet about how lame Apple was because they had a fruit for a logo. Yeah. Right.
What's Up Next?
SGI is headed in a different direction than any other UNIX hardware vendor thus far: SGI has actually embraced Linux and promised significant enhancements, enhancements that may actually bring Linux into the realm of viability for some of us.
IRIX is a rock-solid OS, and I cannot imagine using Linux on a production workstation at this point. But, should some fatherly organization pick Linux up, clean it up, and ensure the interface is one I'm familiar with (ksh and Indigo Magic), and applications suddenly find themselves ported to it, I don't see how it would make much of a difference if I don't notice a performance drop.
Obviously, this will take time. Unlike SGI's impending death any second now (two years and still counting, waiting for the big shoe to drop at any moment), it takes time to truly polish an OS and bring it in line with a production effort. My business is 3D graphics. I work with huge files, huge scenes, and immense amounts of data. The aplomb that an SGI running IRIX demonstrates on my workload is unmatched by any other platform or OS, so much so that I have declined purchasing 'faster' PCs and Macintoshes in favor of Indys and Indigos, simply because the responsiveness wasn't there on the PC/Mac hardware with my typical workload tests.
But if Linux can be brought in line on the low-end on SGI hardware with a comparable SGI running IRIX today, and the only real way of telling the difference is to run uname -a at the prompt, who will care?
Certainly not me, or my work.
Hope this helps.
--
Xenu loves you!
Merill Lynch may know shitloads about stock and what they think the company is doing in the long run, but just because a stock is labelled a "strong sell" doesn't mean the company sucks. From a technical standpoint (the one that, I'd imagine, most /.'ers CARE about), SGI is long from dead. That's what the author is saying. And I completely agree with him. I work in a neuroscience department that uses O2's, Indys, and a challenge for processing huge amounts of image data and dumps statistics based on image analysis and some discrete-logic input. The kinds of simulations they run wouldn't run on a PC or Mac, period. The only thing we get PCs for are sticking the output into a LaTeX or Word document to share with others. I love SGI and what they're doing for the community, as well as the products they've already developed on their own. I don't give a flip what Dean Stanley thinks about them as a business prospect, because I know they'll be around long enough for me. -Chris
>The notion that SGI is "going out of business"
>is certainly valid - the company has dropped key
>divisions and has laid off employees. Added to
>which, the best people left long ago.
Just as happened to IBM ten years ago, and six
years ago, and 3 yeras ago. Obviously signs of
a faltering and soo-to-die company. Obviously
it can't have naything to do with streamlining
operations and concentrating on core competencies.
The same sort of thing has happened at dozens of
high profile computer companies at many times in
the past and no doubt will do so in the future
too.
'The best people' is a purely subjective analysis,
which it's hard to refute. but then it's so vage
it's hard to take seriously either.
>This isn't flamebait - there is a reason why people pay Merrill Lynch and Goldman Sachs to
>do intelligent research. What you see above sure ain't it.
And your post is?
Simon Hibbs
Check out
this page if you're interested in SGI financials. There's been a serious downturn in profitability between 1995 and 1998 which is seriously disconcerting.
I think a potential strategy of integrating the best of IRIX into Linux would be a winner for SGI. Most people aren't buying SGIs for IRIX, are they? So unloading some of that development cost to the Linux community would be a wise move for them, and making XFS a fully-supported file system for Linux means working between the two is easier.
Ooh, a sarcasm detector. Oh, that's a real useful invention.
I've spent six months with an O2000 suffering from irreproducable bus errors, segmentation faults, crashes and CPU hang-ups. SGI has not been able to fix it. It's a level of instablity orders beyond what I ever experienced with Cray (pre-SGI) or any other UNIX.
We have had this problem too, I think I might be able to pull up a log by this afternoon to post if you want to see it. Haven't fixed it yet :-( anyone that knows a solution, PLEASE let me know. I know we lost overnight numbercrunching data at least twice.
The recent SGI announcments are reasonable and I think that their strategy is sound. The big questions is : who's left at SGI? With the 'slightly' different job market for graphics hardware designers and the strange decisions of SGI in the past few years I'm worried that they have lost most of their best engineers, and SGI is nothing without their engineers.
On another note : 88 million polys/s on IR? Where does this come from? An IR pipe does 10 million polys/s AFAIK. The big benefit is bandwidth and the fact that basically all of OpenGL is implemented in hardware (a glVertex3f call turns into 7 (!) instructions on an IR machine), offloading the CPUs to do other stuff. But SGI must do something radical to keep hold of high-end graphics, recent delays in the future graphics line does little to inspire confidence.
I mostly agree with Scott Elyard. But as a SGI user who does not work
in high-end graphics domain, I must say that:
2. SGI has a gloomy future. (SGI is dying, losing all its customers &c.)
I do believe this. In my domain (bioinformatics) which is
fast expanding, less and less people are buying SGI servers and more
and more are buying Sun's and Compaq's. If you don't do high-end
graphics, you really can get better deals elsewhere.
3. IRIX is being abandoned.
This is almost true. People who use IRIX like me get pissed off when
they try to install some new software. Look at setiathome for
instance. It was available for Win32, MacOS, Linux, SunOS. And
that's about it. It took months before an IRIX version was made. So
by merging Linux and IRIX, SGI hopes to make much more software
available on their machines.
Having said that, I am a happy SGI customer. And anyway, I am stuck
with it for a couple of years at least. Fortunately, I managed to
switch my SGI O2 for a Linux PC!
I'll do it for cheesy poofs.
IRIX is a rock-solid OS, and I cannot imagine using Linux on a production workstation at this point.
I have used IRIX since 1993. I have used Linux since 1997. In January of 1998, I might have agreed. As for the whole calander year of 1999, I have spent about 85% more time using Linux, and about 4% of my time using IRIX (and about 0.1% using DEC-UNIX), and about 10% of my time using FreeBSD. Of the time I spend trying to track bugs, and figure out where errors are comeing from, it's about 50/50 IRIX/Linux. Considering how little I use IRIX, that's probably more time tracking bugs than actual use.
To tell the truth, I now 4 people here who use IRIX as thier primary workstation OS, and they would opt out if they had a choice. Conclusing, I think today in 1999, I would completely dissagree with the statement "IRIX is a rock-solid OS, and I cannot imagine using Linux on a production workstation at this point."
The older SGI's we have here without IRIX updated run fine, 100% stable (including a Personal Iris, still going strong). Indy's got a little funny when we went 5.x to 6.x, but are OK now. Indigo is fine. Octane (one of the newest we have) is doing so-so, no problems yet. Origin 200 has been a nightmear. So, it's hit and miss. If you want relyable, I guess you need to look at the older ones, and older versions of IRIX... I think I agree there is a trend that SGI's hardware and software quality is going downhill.
But I am not sure I would go as far as as the poster did.... Well, maybe, but it's hard not to like SGI, like your first true love or something. I guess I don't think at this point I would choose to buy one personally if I were out shopping for something in thier price range for my own, but once upon a time, I would have LOVED to own one.
But a lot of what you are saying is just plain wrong. For starters the summary you link to was posted to the friends of performer mailing list by an SGI employee - and you expect objectivity from that source?
Also, you disagree that SGI's graphic solutions aren't cost-effective compared to consumer 3D boards - even SGI wouldn't risk such as statement (_they_ would realize it would make them look stupid). SGI's target market is only a tiny fraction of the consumer market in terms of the number of products sold, so there's a good reason for this.
But SGI's product is not only being superceded in terms of cost effectiveness. SGI's base reality and Infinite Reality (2) systems are not that impressive compared to the consumer boards. The Nvidia GeForce even has a higher polyrate than the single-pipe IR2, and it certainly has a very comparable fill rate (faster or slower depending on the number of raster managers).
There's nothing that drives product development as fast as the consumer market competion. SGI has faced this reality, why wont you?
What SGI has, but you fail to mention at all (perhaps because it is a _valid_ point), is slightly more features, like 3D textures and multi-sampling (for e.g. anti-aliasing).
Now, about the use of SGI worksstations for making special effects in movies. This has nothing to do with SGI hardware at all - Computer animations in movies are based on raytracing techniques, which to my knowledge are done entirely in software. Nobody is disputing that SGI has the expertise when it concerns graphics - but many are (correctly, imho) disputing that SGIs hardware will be able to stay ahead of the consumer products unless they do something radical. SGI sees the same thing and have formed a strategic alliance with the 3D consumer market leader Nvidia - SGI will built their future hardware using the same chips as consumer boards, only SGI will put them to work in parallel, like some company did with the Voodoo(2) chipset
Linux is not an alternative to SGI yet - the OpenGL hardware support is not solid enough yet, but we're getting there, partially thanks to SGI! Additionally the software isn't there yet either, but we're getting there. For instance SGI(!) is porting performer to linux!
SGI is a cool company with a _lot_ of engineering talent. They really don't need you to spread FUD - Now that the bosses have woken up and demonstrated that they do have the guts to make truly radical changes like adopting linux as a key element in their strategy, they'll make it just fine.
This is an interesting discussion.
Yesterday I was talking with our VP of engineering who came from SGI (where I think he worked on the IRIX OS). He had several interesting comments about SGI.
1. Most of the operating system engineers who worked on IRIX have left. SGI is turning to Linux out of desparation since so many of the people who are knowlegable about the internal workings of IRIX are gone. Apparently SGI's deal with Micro$oft did not go over well with their development team.
2. XFS will be a nightmare to make stable for Linux. XFS is not a clean file system like JFS, but contains a horrendous amount of complexity (read bloat), such as lots of multimedia support, special scheduling code, and so forth. It will be hard to separate the bloat from the core of the file system. Due to its complexity it will be very difficult to understand and make stable. It is likely that SGI is releasing it as open source since they don't know what else to do with it since all of the original developers left.
3. SGI's deal with Microsoft was a huge mistake. Any company that teams with Microsoft is doomed to failure. Microsoft used the deal to essentially kill OpenGL as a valid 3-D platform for Windows and replaced it with Direct-3D.
4. SGI's acquisition of Cray was a huge mistake. The folks at Cray and the Mips division clashed and essentially killed Mips. The Cray folks pushed for vector processing in the Mips processor, which was not a good match. The Mips people argued against this. The Cray team won. The reason is that Cray's bread and butter is their SIMD vector processing. Without it, Cray is nothing. The vector processing conflicted with the whole idea behind RISC, which is to keep things simple. Guess what happened to Mips?
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
As for secure, all you need to do is search the bugtraq archives from '97 and '98 -- nearly every setuid 0 binary had a root-giving exploit for it. I'll agree that any OS can be insecure, sure, but, Christ, IRIX was *really fucking bad* a year or so ago. To its credit, it hasn't appeared much recently on bugtraq/rootshell/etc, so maybe things have finally turned around.
4Dwm still sucks, though.
- A.P. (running WMaker on this IP22)
--
"One World, one Web, one Program" - Microsoft promotional ad
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
>Now, about the use of SGI worksstations for >making special effects in movies. This has nothing >to do with SGI hardware at
> all - Computer animations in >movies are based on raytracing techniques, which >to my knowledge are done entirely in
> software.
Wrong! Many movies are composited (ie the different layers of the animation are added together) on software such Discreet's Inferno system. This runs on
an Onyx2 IR and renders final frames using the OpenGL graphics pipeline. ILM uses it in house but calls it Sabre. Check out www.discreet.com for info. In this market SGI has NO competition, if they ain't using discreet they're using Kodak Cineon or Avid Illusion or Jaleo or Chalice, but they all run on SGI hardware. (yes there is NT compositing solutions but not at the high end).
Also in 3d modeling, sure you render the final frames using software only but during modelling you need a good opengl pipeline so you can see your 50,000 poly texture heavy model at a decent speed to make changes. Try running Maya or Softimage on your TNT or Voodoo 2, what's that, won't run? Funny that.... (Games 3D cards not equal good professional 3d card)
(a visual fx operator and everyday sgi user)
as an sgi admin user and long time sgi fan - i'd like to point out that SGI has the crappiest licensing scheme of *any* unix out there - nodelocked licensing sucks. are you on crack SGI ? Godammit if im running a critical app like proe and my license machine goes down, the whole department is shut down from using proe..the backup servers dont seem to do their job - GET YOUR HEAD OUTA YOUR ARSE SGI. Expensive licensing, modelocks, time limited stuff, a compiler (MIPS Pro) thats basically incompatible with GCC (slightly different language syntax - why ? why ?)
Irix has been going downhill since 5.3..6.5 sucks..cant even do NFS properly.
But, GCC is no where near as good a compiler is MIPS Pro 7.3 for thier hardware.
It "dosen't?"
?ttp://finance.yahoo.com/q?s=SGI&d=my Look at graph. Ford, GM, even Dodge doesn't do this (unless they are following the rest of the market). SGI's droping while other Tech stocks rise.