Halting the 320 & 540 was the best decision SGI made in relation to the whole NT thing.
Sad to say, I have to agree with you. I loved the 320; I came close to buying one for my home (and cash flow is tight with another kidling on the way). At the same time, I need SGI to stay in business; they're still making the best tools for my work. If SGI can manage to make something much like the 320/540 but is commercially viable over the long haul and runs all my favorite applications (!), I'm there.
Re: MIPS = dead: somebody's confused. MIPS makes the most popular embedded processor in the world. Period. Maybe User X (who may or may not have worked for SGI in the past) doesn't do floating point calcs anymore, but 3.5 hours for a SETI work unit is a serious motivator in my business. Signal processing and visualization at the same time?!;-)
BTW, when I heard about the "SGI is a UNIX company = Linux + IRIX" press release, my first response was, "Now there's a ballsy decision." I still feel it was the right move (and I wish I had had serious money to buy SGI when it was at $6 per share).
PS: Let me know if SGI could use any east coast industry-advocates. I accept payment in cool t-shirts....;-)
[An early 1990s beta-tester for Crimson, Irix 5.0a, Delta C++, etc.]
This is actually a very insightful take on SGI's current strategy.
Thanks, I like pretending I know what I'm talking about....;-)
The "NT mindshare" question is an interesting one, as is the "move." As far as I can tell, smart folks are thinking "we want to move away from HP (or VAX or whatever) hardware to something that's more commercial / consumer grade." The smart ones are not saying, "NT is better than UNIX." They just haven't had a viable choice up until now (remember, all decisions must be defensible).
I am literally watching government project managers changing their minds and saying, "No, we've decided to go with Linux rather than NT on our PC-based-replacement for legacy hardware. Why? Because it's X-Windows, Unix-ish, etc., and thus has lower risk than MFC-ing, NutCrackering or Java-ing our code." We're seeing the NT "pull" evaporate; as far as I can tell, the only "push" left is inertia (of which the government has a lot).
Two somewhat related items:
1. I'd seen several remarks implying that SGI hardware rules SETI but might have limited utility elsewhere. I think we've agreed that if something is in use (and has strong advocates), it's gotta have measurable utility.
However, I'd never tried the SETI work units on my R12000 at work. Given that some benchmarks indicate that Pentium III-based machines are right on top of SGI performance (e.g., CPU95 at www.spec.org), I thought it'd be interesting to try a "real" cross-platform application (remember, your mileage may vary; chances are slim that your full-time employment is as a SETI scientist...):
Swell, the R12000 does great FFTs, what's the point? The thing is, that's the kind of thing I do: lots of real-time floating point. More importantly, we're not talking about a few percent; this is a different order of magnitude! For completeness, compare the price on a Dell workstation (as tested for SPEC) to a single R12K SGI Octane. The SGI will likely be more but not by an order of magnitude....
2. In terms of "pennywise", I've started pushing for the idea that if we developers are willing to have less people (at hiring time, not layoffs!), the management should reinvest that "savings" into our infrastructure. Salary + benefits + overhead comes to quite a lot in today's economy. Split between a group of 3-4 developers, it can buy lots and lots of toys... I mean "tools"!;-) Seriously, though, that's also an annual savings; keep that infrastructure up to date.
I've actually started using it as a metric for employment; if your PHBs just pocket the money, you know it's time to look for another job.
Well, I wasn't directing my remarks at you personally, but you do raise a point that has always baffled me: Linux will run on low cost, high powered hardware. IRIX will (currently) only run on expensive proprietary SGI boxes.
I can easily put together a set of requirements that would force a Linux-based system to cost more than the "equivalent" SGI box. Real-time image processing on a video stream is one example: an R5000 O2 can do a lot in that regime (for tiny money); I don't know how you'd build the Linux box to do the same thing (assuming PC components).
As I said before, I'm a big fan of developer-centric tools (or any type of worker). The worker costs money: getting them the proper tools is always a win (if not by increasing profit margins then by increasing volume of work, thus lowering the need for higher profit margins). If expensive worker X needs a tool that requires a dedicated machine running BeOS to do something that must be done, the decision is obvious. Of course, nobody needs to believe me, I have the wrong letters after my name to be talking about return on investment, net present value, total cost of ownership, etc. Ask my boss, he's an MBA.;-)
The funny thing is, SGI is smart. They know that if they help drive the Linux application market, they will reap the benefits: "Hey applications maker, let us help you port your Linux application to Irix."
Here's the punchline: if SGI is starting to get rolling in terms of supporting the Linux / Open Source community, doesn't that make them a Linux start-up? Shouldn't their stock price be doing a RedHat?;-)
I have heard it said, and I will say it myself, Irix is perhaps the BEST operating system in existence right now.
Well, I'm definitely in agreement with a qualifier similar to your other comment: It is a joy to be a developer on IRIX.
That's really the key isn't it: development. What people forget that the most expensive resource isn't the computer; it's the developer. Chances are very slim (in today's economy) that, if you're a working developer, you cost your employer less than the computer that you work at. Even better, your fully loaded rate (with overhead, etc.) is likely several times higher than an amortized billing / lease rate for your machine.
In short, it makes economic sense to pick the tools that make your (expensive!) workers most effective. In my case, I've been working with UNIX (and other OSes) for about 20 years; I find IRIX to be the best all-around package. Then there's the real-time capabilities: nanosecond-level real-time clocks, anyone?
On the other hand, for those running render farms (batch processing of animations), they need cheap cycles, so they don't really care about "development." To my mind, they shouldn't be buying computers at all, they should be renting the cycles. That sort of drives down your "total cost of ownership...".;-)
My favorite comments, though, are when person P says something along the lines of "application X isn't supported on IRIX, so it sucks and is dying off" (e.g., Multigen Creator).
Well, if application X is a requirement right now, it's a requirement; choose a supported a platform and get to work. Note, however, that a lot of those types of applications are also not supported on Linux. Well, P, by the transitive property, you're saying that Linux sucks and is dying off, too. Are you sure you want to say that? Here? Go on, I dare you....;-)
You know, I appreciate the innovation and development that goes into consumer graphics boards. Hell, my dissertation was based on the assumption that "soon all computers will be as powerful as the RumptyRump BFG 9000."
But
We've had a word for statements like "$100 graphics chips which outperform $100,000 Silicon Graphics workstations" for about 20 years now: it's "troll."
For clarity, I've worked with $200+K SGIs that are now selling used for very low prices (and that's a good thing). With all the extra stuff, the machine on my desk went for $60K (SGI Octane). Would I replace it with an "equivalent" PC-based product? Don't be absurd.
For those responding with "hey, why not?", look up "crossbar switch", "gigaflop", "multiprocessor" and "multiple asynchronous data sources."
PCs are not workstations - and that's okay! Right now, I'm typing from my home machine: a Pentium-based PC with an NVidia card. Am I most effective from this machine? No. Is it okay for home use (slashdot, web surf, games, games, games and a bit a surplus work)? Sure.
Am I worth more to my employer banging on big iron (as opposed to something with "$100 graphics chips")? Apparently so.... $-)
SGI's GL pipeline is its raison d'etre. If developers stop using it, SGI doesn't exist.
I definitely agree and I'll underline it by saying something more general: the reason SGI hasn't died off is because developers continue to use SGI Whatever (be it hardware or software). Remember, the most expensive item in any project is the labor: use the tools that help your developers work fast and well.
This is not about games, this is about taking back/retaining the CAD and visualization markets.
I also agree that this is a big factor but let me put a different spin on it. Say there's a defense contractor who makes his living doing OpenGL programming on government projects (hm, who could he be thinking about;-). He, a Certified SuperGenius, determines that there is a growing market for Linux-based games, especially graphically intense ones (there is, right?;-), in which a reinvestment of graphics smarts would really pay off. But, the company's current core competency is deep but not broad: OpenGL, Performer, UNIX, real-time simulation, etc. What would motivate that guy to go to his boss and say, "Hey, we need to start working on this. Linux isn't necessarily there yet, but it will be"?
Commitment to standards by the players (or at least some of the players; if there's one, there will be more). We call that risk management: specifically, management of technical and staff risk. "But boss, we already know how to do this part, so we don't have to worry about it. We can concentrate on the new stuff without worrying about our foundation collapsing."
Two points above:
1. It's not just the CAD, etc. market, though that's important.
2. Any technical innovations are good for Linux and the market in general, usually for reasons you didn't expect.
Of course, now I have to shake lose some game development funding....;-)
I've seen quite a bit of confusion on/. about Performer, why it's a good thing, why is SGI so stupid / brilliant, etc. so I figured that I'd summarize some of the key points that are most important to me as a graphics guy who does a lot of graphics-newbie indoctrination. For true Performer-heads reading, remember that I'm being purposely "high level", so feel free to add detail that I'm glossing over.
Background: I'm an SGI hack from way back (not all the way back, but close) with lots of hours logged on everything except their most recent Origins and Onyx2's. I was migrating my IrisGL (OpenGL hadn't been invented yet) code from C to C++ when the original Performer 1.0 framework starting wandering out of the labs. Since then, it's much more OOD - OpenGL and C++ have greatly increased its usefulness without over "object"ing it.
Here are some of the things that Performer can give you quickly (i.e., not much more complex than Performer "Hello World"):
A shallow learning curve: from "Hello World" to something that looks impressive is a very quick process. For example, somebody hands you an arbitrary ".obj" (or whatever) object file and says "I want this in the virtual environment." You write less than ten lines of code to create a new object, point its geometry descriptor at the model (i.e., type in the filename) and recompile. Done. Of course, you haven't made it move yet; that's another line of code.
A framework for quickly assembling a visual simulation / application. Example: somebody hands you some 3D models and you want to make them fly around each other. Instantiate some objects, point their geometry descriptors at the model files, write some simple "move me around" code and you're done. No OpenGL, no worrying about clipping at the edge of the screen - it's already taken care of.
A platform independent framework - your Linux Performer code will compile on my Irix machine and, assuming 2.3 and 2.2 aren't too diverse, vice versa. Even more interesting, within the Irix world, a Performer app scales across the hardware spectrum. If there's a bunch of processors, it becomes a multiprocessor app at runtime. No recompiles between O2, Octane, IR, etc. Obviously, if your simulation includes big number crunching or huge graphics loads, it will have trouble fitting in a small box but at least it will run at all. I don't know how this translates to the Linux world - I don't have the hardware suite to do a good test....
[This is getting long so I'll wrap up.]
Why does Slashdot and the rest of the Linux crowd care? Well, the premier all-around computer graphics company is handing you their flagship visual simulation framework and saying, "we hope you have a good time." The marketability of Linux just went up by orders of magnitude - simple example in my field, it counters HP's arguments that they're a better buy for military simulations. "Gosh, you're more expensive than Linux / IRIX and not as powerful. Why exactly are you better?"
Why does SGI care? The one place they've always lost is marketing - in a word, they stink at it. They need the groundswell from the popular marketplace. So, be loud. If you try it and like it, say so. If you find things that you need or don't work, complain loudly and constructively.
Also: be quotable. Make sure that it's very obvious that you're a *nix-head running SGI software. Give them the ammunition and SGI will produce high-coolness useful power for all of us.
I feel that the release / announcement of _new_ SGI Linux boxes is a good thing because it was something planned some time ago; thus, even though they've had a change of leadership, they're still committed to this as a path. Personally, I think they've made a good decision: "we're a UNIX company and we're good at it, we'd never win by trying to fight the NT fight."
My view on the SGI Linux effort on Intel is that they're trying to do exactly what's been mentioned here: grab mindshare (and thereby marketshare).
This is the image that they want to put in your mind: a spectrum of machines, rangine from your desktop (you're running Linux on your desktop, right?;-), through file & compute servers and graphics workstations all the way up to monster multi-processor visualization and GP computing beasts. The interesting thing is that some of those machines would be running Linux on Intel, some IRIX on MIPS - they want those lines to blur. Even better, the machines could be source level compatible - not quite as sweet as a MIPS-only shop (just drop the binaries anywhere), but just a recompile away....;-)
Finally, I don't think SGI is spending money on Linux on MIPS because they understand the Linux community: if the Linux community at large really really wanted Linux on MIPS, they'd make it. If there were a loud movement that said "yes, we do want it but we need more information to make it sweet", I could see them coming across. Open Source = good business for them.
Finally finally;-), I'd bet they'd dearly love to see their Intel-based Visual Workstations running Linux with full hardware-supported graphics (last I checked, it could only do software-based). Any takers...?
On the IRIX side, we have many of the SGI-specific extensions that bring OpenGL up to par with the original top-flight IrisGL spec (Onyx RealityEngine2 was the last time I saw that bar go up). I'd like to see every one of those extensions move into the mainstream, even if they're only supported in hardware (check downstream of www.sgi.com if you're curious about what I talking about).
My personal favorite is one that you never ever hear about because, as far as I know, no PC graphics solutions have ever made it work: multisampling. Def: hardware supported, full-framebuffer antialiasing using (get this, my favorite part) _one_ extra line of code.
Basically, the big boxes allowed you to use a small primary display area multiplied several times throughout the (not visible) framebuffer - each rendered at a slightly different viewpoint. These images were then all composited in hardware into the display area for each frame before showing up to you as a beautifully smooth image (I generally used 4 samples per pixel, sometimes higher than 120 frames per second). And, for other graphics guys, this is not the same as the accumulation buffer - this all took place at the hardware level, no programming required.
Why is this relevant to gaming? Well, hell, if you could have a jaggie-free display without having to look up (or debug!) a single antialiasing algorithm, wouldn't that be good and stuff?;-)
Are we likely to see this on a PC board tomorrow? Probably not. Is there any reason to _not_ include the necessary entry points in the standard OpenGL spec? Likewise, probably not. Eventually hardware will catch up (I hope).
I think we're in violent agreement, to a certain extent. But:
> Heck, for the sake of fairness, spot them the OS, so that both teams have the same hardware budget.
I still don't believe that this is "fair." Benchmarks by nature try to convince you that their arbitrary imaginary scenario is fair. What I'm trying to propose is that context _matters_: one should start from a _real_ problem. For example, "the boss gave me a budget of $X, a set of requirements and told me that it's my ass on the line if the solution I concoct doesn't deliver" (or goes over budget - e.g., due to unforseen training needs).
Note that I'm not proposing that one or the other would win: I don't have enough personal information to make a reasoned hypothesis (I'd place a bet but not a bid). All I'm asking for is for some of the serious architects out there in Slashdot land to do the math and show how things would really stand up if we were playing for real money.
All that being said, I still get pumped when SGI posts a huge SPECfp-rate number (http://www.spec.org) - still my OS of choice for Big Problems(TM) $-).
And I hear my Octane is in the building now....;-)
Actually, I've been waiting for the argument that seems to apply most directly: that benchmarks don't really represent the Real World(TM). For example, given the oft-cited media demographics of the "poor little pre-IPOs who just don't have enough money to buy M$'s latest work-in-progress," you'd think there would be some tracking of multi-function boxes.
For example, if you're using Linux as a major player at work, do you have a _pure_ web server box and a _pure_ file server box? Or do you have one or more boxes that really do lots of different stuff all the time?
This is a totally different measurement and gets directly to the heart of the "Bang for the Buck" argument: if I have $K dollars to spend on _one_ box to do everything, would I rather spend it on the best hardware (with cheap, powerful, easy to use) that I can get? Or am I going to have to spread it around on hardware+expensive OS?
At this point, you're talking real "business case" numbers that you can talk to your boss / capital approver and say "See, the path I recommend makes more sense in terms of total functionality, operator training, total cost of ownership, blah blah blah". (Anybody who's ever had to go through that drill knows what to put there - I just justified a $60K IRIX box for my desk... 8-).
So, the benchmarks that I'd really like to see are things like the following:
1. Given $5K, $10K or $20K to spend on a _single_ general purpose server machine, inclusive of OS and any serving technology, what's the best of class? Linux should win here by definition: cheap OS = big box, expensive OS = littler box.
2. Given the test platforms from #1 above, run through the Mindcrafty benchmarks again. At this point you're comparing dollars to dollars, so the results might be interesting this time. Not useful (steady state measurements don't indicate real life - does the Slashdot effect slowly ramp up or hit all at once...;-).
3. Now show me something interesting. E.g., X web clients connected, file server traffic jumps up, then web clients drop off. You know, scientific method and all that?
At this point, I'd say the Mindcraft data is still far from being useful. Simply put, there's not enough of it.
Re: thanks - no sweat. It's what we professoresque types do....;-)
Re: the non-compete, it's unlikely that it's enforceable / relevant outside of VA (maybe not even _in_ VA). The IP ownership definitely is, though, so don't play that game if you can avoid it. In the future, one of the best ways to defend yourself from this tactic (trying to steal your pre-job school-days ideas) is to publish _anywhere_. You don't have to put in every last detail if you think it's going to be marketable, but it proves prior art in one step ("as you can see, I was writing about this four years ago..."). Most CS departments have something like Technical Reports, if you don't want to try for something higher order (conference, journal or magazine) - a pretty easy place to make things a matter of record: here's what I've been working on, here's how it's going, here's where it's going, etc.
Final thought on the whole "those bastards are stealing my ideas" problem: do you really think it's likely that they'll be able to do a thing with them since you're leaving? I've had to leave code behind before and I get a big kick out of hearing from my spies about how the PHB just kind of stared at it: "Duh, what's this do?" "That's an rgbColor. You know, red, green, blue?" "Uh, yeah!" Yeah, best of luck with that, weenie boy!:-p
Re: grad school - there's some really sweet graphics / VR / visualization schools out there these days: GA Tech is still #1 (unless I've been out of the loop long enough), Brown #2. Other examples: UNC Chapel Hill, UVA, Utah, Indiana Univ (go figure - they have a load of hardware, though, thanks to the efforts of former grad students;-). Discussion of the ins and outs of grad school is a different thread, though....
This is a bit spooky. I've lived this life: same place, different details. I'll spill the beans on how the story turns out: paying dues early = victory in the long term (now define "victory"...;-).
More comments below:
> For the on-topic bit: I was suckered out to Virginia for $50k/year > fresh out of college, but I've got lots of skill in realtime 3D > which this company wanted, but they didn't know how to treat their > employees and so I left after a few months.
I feel your pain. Note, however, that the little box on an HR drone's checklist can now be checked: "Have you worked in your field outside of college?" You also have a decent price point that you could use in future negotiations, especially if you're planning on relocating: $50K in Virginia dollars is worth less than $50K in Teeny Tiny New England state's dollars. I've definitely done that math.
> For the off-topic bit: I have come to realize that the industry > isn't for me. Academia is where I belong. I'm not a mercenary > programmer. So of course, after a few months of living relatively > large (figuring I'd be gainfully employed for a long time) I'm > having my world kinda crash down around me, financially anyway. It > doesn't help at all that I incurred some debt in moving out here > which I, very stupidly, put off paying back. All in all, I'd have > about broken even for the whole experience were it not for the > various tech toys I suddenly found myself able to buy... Even though > I rationally know that grad school is best for me, and emotionally > know it as well, it just doesn't help to have all you mercenary > types rubbing my nose in what kinds of salary I'm giving up.:)
This is where the popular culture seems to break down: the "industry vs academia" dichotomy is really a spectrum. There's lots of jobs that fall in between the two extremes: in my case, since last graduation, I've been a post-doc, a "Scientist," a "Manager" and now (just recently) a "Senior Scientist." Salary is more than double what it was in VA, I'm off the evil "100% management" track (though I've checked some of those boxes on those aforementioned HR sheets) and I'm back into pretty serious R&D in my field (real-time 3D graphics - see, spooky!;-). Kind of like a professor, but I don't have to teach or grade papers (hooray hooray!;-).
> I've never been into computing and programming for the money, except > for a brief period of time when I was graduating college and I got > suckered into putting off my happiness for the promises of getting > to keep doing the cool stuff while also making enough money to live > very comfortably. Of course, those promises never panned out, and > the company I got hired by turned out to be nothing more than a pair > of two-bit swindlers doing whatever they could to control spineless > employees who didn't know better and weren't at liberty to leave for > a variety of reasons.
Again, I feel your pain. The suck factor tends to be pretty high after college / university. The trick is to be constantly thinking about where you're going, not just where you are, especially as the suck factor tends towards maximum. Consider your next job interview:
1. You've worked in your field and you've set a salary point. 2. You've learned about how to do things _wrong_ in the business place. If you can talk about this effectively (i.e., unemotionally) with the next employer, you start sounding like a "seasoned veteran." Obviously, it takes more than one job to get all that seasoning, but still. 3. You're in a slick job market. \begin{irony} If you stay in VA, there's only about five graphics people in the whole state and they just seem to keep moving from job to job, doubling their salary. Of course, the reason the supply is low is that the environment stinks, but hey, money is god, right?;-) \end{irony}
Likewise, I seem to be the only graphics guy in this little Teeny Tiny State (see above), so the last time I was on the market, I had a jumbo offer in a week. Of course, I turned that one down because they tried to hit me with a "don't work in your field for a year" non-compete, it wasn't really in my field / area of interest, and the commute sucked!
Instead, I'm working directly in my field, doing the kinds of things I'd want to be doing in the university (if I didn't have to teach / get tenure), the company just bought me a brand-new great big machine (costs _way_ more than double the price of my car $-) and I'm working with people that don't suck. Sounds about right to me!
> I need to give some advice to academic types who might be reading > this thread: which do you prefer, money or happiness?
Summary: it doesn't have to be an XOR. So sayeth the Academic Type.
> I used to process all the Security Clearances in Pacific Region in > Canada, as well as declassify a lot of information in personnel > files. And held a Secret clearance.
I'm on the opposite of the coin, re-activating my clearance again to get back into classified projects (again). I've been dutifully re-reading my briefing materials, especially about producing classified documents, and have found them to be pretty up-front and straightforward. Here's a summary (a techie summary, not an "executive" summary) for those "not in the know":
1. Classified data = _real_ data. Approximations show up all over. If you're including actual dates, frequencies, or multiple decimal points, you're probably producing something that's >= Secret.
2. One classified number classifies that paragraph and, therefore, one or more paragraphs, depending on your writing (notice, we're talking paper here). Other pages may be unclassified (e.g., no reeeally accurate numbers), but a document including the SuperDuper Secret pages is SuperDuper Secret by the transitive property.
3. You are allowed (and encouraged) to produce unclassified documents / reports if at all possible. In the above example, take out the really neat number and your paragraph, page and document level drop to lowest possible (e.g., Unclassified).
4. Classification is not a toy (WillAffleck's example is depressing but probably common). The documented government policy here is "don't be a dick."
5. Classification exists for a reason. The best example I've ever heard was when talking with mumble mumble about pre/during-Vietnam War adventures and why he _still_ wouldn't talk about real details, his reply was "They(TM) didn't know we could get a submarine in there." I.e., if they didn't know we could get there, they don't know _why_ We(TM) were there and what We(TM) were there for. Think about it, when someone goes looking for something, it's because someone told them "there's something to look for here." Those are some of the "assets" that are protected by TopSecret++.
> Most, 95%, of the material classified as Secret is junk. The same > holds for Confidential. I presume, based on inference, that probably > 50% of Top Secret material is junk.
6. [Partial agreement] Confidential is silly: "potential risk to national security"?! [I'm quoting poorly but I don't have the definition in front of me.] The Jerry Springer show is a known risk to national security (it decreases the average IQ of the nation ==> bad ==> national security-- ), should it be classified Secret?
Blah blah blah - this was a bit more than I intended but sometimes a definition of terms is useful.
The punchline is: sometimes (often) declassification is a good thing; just getting the paperwork processed decreases the entropy / bureacracy of the universe. Sometimes, declassification is literally not possible: in extreme cases, the "asset" in question is still in "use" and we really don't want people to know how far the "asset" can shoot, or where the "asset" lives / works / sends his/her children to school, etc.
Finally, the easiest way to tell if somebody is working on something classified is if every answer is a variant of "I don't remember." I still play this game with mumble mumble above:
DB: "So where'd you go in that sub?" MM: "Out to sea." DB: "For how long?" MM: "A while." DB: "How long could you stay submerged?" MM: "Long enough."
== Doctor Bob Clearance: I don't remember right now. Project: Yes.
Slashdot Mirror
Halting the 320 & 540 was the best decision SGI made in relation to the whole NT thing.
Sad to say, I have to agree with you. I loved the 320; I came close to buying one for my home (and cash flow is tight with another kidling on the way). At the same time, I need SGI to stay in business; they're still making the best tools for my work. If SGI can manage to make something much like the 320/540 but is commercially viable over the long haul and runs all my favorite applications (!), I'm there.
Re: MIPS = dead: somebody's confused. MIPS makes the most popular embedded processor in the world. Period. Maybe User X (who may or may not have worked for SGI in the past) doesn't do floating point calcs anymore, but 3.5 hours for a SETI work unit is a serious motivator in my business. Signal processing and visualization at the same time?!
BTW, when I heard about the "SGI is a UNIX company = Linux + IRIX" press release, my first response was, "Now there's a ballsy decision." I still feel it was the right move (and I wish I had had serious money to buy SGI when it was at $6 per share).
PS: Let me know if SGI could use any east coast industry-advocates. I accept payment in cool t-shirts....
[An early 1990s beta-tester for Crimson, Irix 5.0a, Delta C++, etc.]
This is actually a very insightful take on SGI's current strategy.
;-)
... I mean "tools"! ;-) Seriously, though, that's also an annual savings; keep that infrastructure up to date.
Thanks, I like pretending I know what I'm talking about....
The "NT mindshare" question is an interesting one, as is the "move." As far as I can tell, smart folks are thinking "we want to move away from HP (or VAX or whatever) hardware to something that's more commercial / consumer grade." The smart ones are not saying, "NT is better than UNIX." They just haven't had a viable choice up until now (remember, all decisions must be defensible).
I am literally watching government project managers changing their minds and saying, "No, we've decided to go with Linux rather than NT on our PC-based-replacement for legacy hardware. Why? Because it's X-Windows, Unix-ish, etc., and thus has lower risk than MFC-ing, NutCrackering or Java-ing our code." We're seeing the NT "pull" evaporate; as far as I can tell, the only "push" left is inertia (of which the government has a lot).
Two somewhat related items:
1. I'd seen several remarks implying that SGI hardware rules SETI but might have limited utility elsewhere. I think we've agreed that if something is in use (and has strong advocates), it's gotta have measurable utility.
However, I'd never tried the SETI work units on my R12000 at work. Given that some benchmarks indicate that Pentium III-based machines are right on top of SGI performance (e.g., CPU95 at www.spec.org), I thought it'd be interesting to try a "real" cross-platform application (remember, your mileage may vary; chances are slim that your full-time employment is as a SETI scientist...):
3 h 27 m: PIII 500 Mhz Wintendo98 = 4%
16 h 22 m: PMMX 233 Mhz WinNT = 58%
3 h 27 m: R12K 300 Mhz Irix 6.5 = 100%
Swell, the R12000 does great FFTs, what's the point? The thing is, that's the kind of thing I do: lots of real-time floating point. More importantly, we're not talking about a few percent; this is a different order of magnitude! For completeness, compare the price on a Dell workstation (as tested for SPEC) to a single R12K SGI Octane. The SGI will likely be more but not by an order of magnitude....
2. In terms of "pennywise", I've started pushing for the idea that if we developers are willing to have less people (at hiring time, not layoffs!), the management should reinvest that "savings" into our infrastructure. Salary + benefits + overhead comes to quite a lot in today's economy. Split between a group of 3-4 developers, it can buy lots and lots of toys
I've actually started using it as a metric for employment; if your PHBs just pocket the money, you know it's time to look for another job.
Well, I wasn't directing my remarks at you personally, but you do raise a point that has always baffled me:
;-)
;-)
Linux will run on low cost, high powered hardware. IRIX will (currently) only run on expensive proprietary SGI boxes.
I can easily put together a set of requirements that would force a Linux-based system to cost more than the "equivalent" SGI box. Real-time image processing on a video stream is one example: an R5000 O2 can do a lot in that regime (for tiny money); I don't know how you'd build the Linux box to do the same thing (assuming PC components).
As I said before, I'm a big fan of developer-centric tools (or any type of worker). The worker costs money: getting them the proper tools is always a win (if not by increasing profit margins then by increasing volume of work, thus lowering the need for higher profit margins). If expensive worker X needs a tool that requires a dedicated machine running BeOS to do something that must be done, the decision is obvious. Of course, nobody needs to believe me, I have the wrong letters after my name to be talking about return on investment, net present value, total cost of ownership, etc. Ask my boss, he's an MBA.
The funny thing is, SGI is smart. They know that if they help drive the Linux application market, they will reap the benefits: "Hey applications maker, let us help you port your Linux application to Irix."
Here's the punchline: if SGI is starting to get rolling in terms of supporting the Linux / Open Source community, doesn't that make them a Linux start-up? Shouldn't their stock price be doing a RedHat?
I have heard it said, and I will say it myself, Irix is perhaps the BEST operating system in existence right now.
Well, I'm definitely in agreement with a qualifier similar to your other comment:
It is a joy to be a developer on IRIX.
That's really the key isn't it: development. What people forget that the most expensive resource isn't the computer; it's the developer. Chances are very slim (in today's economy) that, if you're a working developer, you cost your employer less than the computer that you work at. Even better, your fully loaded rate (with overhead, etc.) is likely several times higher than an amortized billing / lease rate for your machine.
In short, it makes economic sense to pick the tools that make your (expensive!) workers most effective. In my case, I've been working with UNIX (and other OSes) for about 20 years; I find IRIX to be the best all-around package. Then there's the real-time capabilities: nanosecond-level real-time clocks, anyone?
On the other hand, for those running render farms (batch processing of animations), they need cheap cycles, so they don't really care about "development." To my mind, they shouldn't be buying computers at all, they should be renting the cycles. That sort of drives down your "total cost of ownership...".
My favorite comments, though, are when person P says something along the lines of "application X isn't supported on IRIX, so it sucks and is dying off" (e.g., Multigen Creator).
Well, if application X is a requirement right now, it's a requirement; choose a supported a platform and get to work. Note, however, that a lot of those types of applications are also not supported on Linux. Well, P, by the transitive property, you're saying that Linux sucks and is dying off, too. Are you sure you want to say that? Here? Go on, I dare you....
You know, I appreciate the innovation and development that goes into consumer graphics boards. Hell, my dissertation was based on the assumption that "soon all computers will be as powerful as the RumptyRump BFG 9000."
But
We've had a word for statements like "$100 graphics chips which outperform $100,000 Silicon Graphics workstations" for about 20 years now: it's "troll."
For clarity, I've worked with $200+K SGIs that are now selling used for very low prices (and that's a good thing). With all the extra stuff, the machine on my desk went for $60K (SGI Octane). Would I replace it with an "equivalent" PC-based product? Don't be absurd.
For those responding with "hey, why not?", look up "crossbar switch", "gigaflop", "multiprocessor" and "multiple asynchronous data sources."
PCs are not workstations - and that's okay! Right now, I'm typing from my home machine: a Pentium-based PC with an NVidia card. Am I most effective from this machine? No. Is it okay for home use (slashdot, web surf, games, games, games and a bit a surplus work)? Sure.
Am I worth more to my employer banging on big iron (as opposed to something with "$100 graphics chips")? Apparently so.... $-)
SGI's GL pipeline is its raison d'etre. If developers stop using it, SGI doesn't exist.
;-). He, a Certified SuperGenius, determines that there is a growing market for Linux-based games, especially graphically intense ones (there is, right? ;-), in which a reinvestment of graphics smarts would really pay off. But, the company's current core competency is deep but not broad: OpenGL, Performer, UNIX, real-time simulation, etc. What would motivate that guy to go to his boss and say, "Hey, we need to start working on this. Linux isn't necessarily there yet, but it will be"?
;-)
I definitely agree and I'll underline it by saying something more general: the reason SGI hasn't died off is because developers continue to use SGI Whatever (be it hardware or software). Remember, the most expensive item in any project is the labor: use the tools that help your developers work fast and well.
This is not about games, this is about taking back/retaining the CAD and visualization markets.
I also agree that this is a big factor but let me put a different spin on it. Say there's a defense contractor who makes his living doing OpenGL programming on government projects (hm, who could he be thinking about
Commitment to standards by the players (or at least some of the players; if there's one, there will be more). We call that risk management: specifically, management of technical and staff risk. "But boss, we already know how to do this part, so we don't have to worry about it. We can concentrate on the new stuff without worrying about our foundation collapsing."
Two points above:
1. It's not just the CAD, etc. market, though that's important.
2. Any technical innovations are good for Linux and the market in general, usually for reasons you didn't expect.
Of course, now I have to shake lose some game development funding....
I've seen quite a bit of confusion on /. about Performer, why it's a good thing, why is SGI so stupid / brilliant, etc. so I figured that I'd summarize some of the key points that are most important to me as a graphics guy who does a lot of graphics-newbie indoctrination. For true Performer-heads reading, remember that I'm being purposely "high level", so feel free to add detail that I'm glossing over.
Background: I'm an SGI hack from way back (not all the way back, but close) with lots of hours logged on everything except their most recent Origins and Onyx2's. I was migrating my IrisGL (OpenGL hadn't been invented yet) code from C to C++ when the original Performer 1.0 framework starting wandering out of the labs. Since then, it's much more OOD - OpenGL and C++ have greatly increased its usefulness without over "object"ing it.
Here are some of the things that Performer can give you quickly (i.e., not much more complex than Performer "Hello World"):
[This is getting long so I'll wrap up.]
Why does Slashdot and the rest of the Linux crowd care? Well, the premier all-around computer graphics company is handing you their flagship visual simulation framework and saying, "we hope you have a good time." The marketability of Linux just went up by orders of magnitude - simple example in my field, it counters HP's arguments that they're a better buy for military simulations. "Gosh, you're more expensive than Linux / IRIX and not as powerful. Why exactly are you better?"
Why does SGI care? The one place they've always lost is marketing - in a word, they stink at it. They need the groundswell from the popular marketplace. So, be loud. If you try it and like it, say so. If you find things that you need or don't work, complain loudly and constructively.
Also: be quotable. Make sure that it's very obvious that you're a *nix-head running SGI software. Give them the ammunition and SGI will produce high-coolness useful power for all of us.
I feel that the release / announcement of _new_ SGI Linux boxes is a good thing because it was something planned some time ago; thus, even though they've had a change of leadership, they're still committed to this as a path. Personally, I think they've made a good decision: "we're a UNIX company and we're good at it, we'd never win by trying to fight the NT fight."
;-), through file & compute servers and graphics workstations all the way up to monster multi-processor visualization and GP computing beasts. The interesting thing is that some of those machines would be running Linux on Intel, some IRIX on MIPS - they want those lines to blur. Even better, the machines could be source level compatible - not quite as sweet as a MIPS-only shop (just drop the binaries anywhere), but just a recompile away.... ;-)
;-), I'd bet they'd dearly love to see their Intel-based Visual Workstations running Linux with full hardware-supported graphics (last I checked, it could only do software-based). Any takers...?
My view on the SGI Linux effort on Intel is that they're trying to do exactly what's been mentioned here: grab mindshare (and thereby marketshare).
This is the image that they want to put in your mind: a spectrum of machines, rangine from your desktop (you're running Linux on your desktop, right?
Finally, I don't think SGI is spending money on Linux on MIPS because they understand the Linux community: if the Linux community at large really really wanted Linux on MIPS, they'd make it. If there were a loud movement that said "yes, we do want it but we need more information to make it sweet", I could see them coming across. Open Source = good business for them.
Finally finally
On the IRIX side, we have many of the SGI-specific extensions that bring OpenGL up to par with the original top-flight IrisGL spec (Onyx RealityEngine2 was the last time I saw that bar go up). I'd like to see every one of those extensions move into the mainstream, even if they're only supported in hardware (check downstream of www.sgi.com if you're curious about what I talking about).
;-)
My personal favorite is one that you never ever hear about because, as far as I know, no PC graphics solutions have ever made it work: multisampling. Def: hardware supported, full-framebuffer antialiasing using (get this, my favorite part) _one_ extra line of code.
Basically, the big boxes allowed you to use a small primary display area multiplied several times throughout the (not visible) framebuffer - each rendered at a slightly different viewpoint. These images were then all composited in hardware into the display area for each frame before showing up to you as a beautifully smooth image (I generally used 4 samples per pixel, sometimes higher than 120 frames per second). And, for other graphics guys, this is not the same as the accumulation buffer - this all took place at the hardware level, no programming required.
Why is this relevant to gaming? Well, hell, if you could have a jaggie-free display without having to look up (or debug!) a single antialiasing algorithm, wouldn't that be good and stuff?
Are we likely to see this on a PC board tomorrow? Probably not. Is there any reason to _not_ include the necessary entry points in the standard OpenGL spec? Likewise, probably not. Eventually hardware will catch up (I hope).
I think we're in violent agreement, to a certain extent. But:
;-)
> Heck, for the sake of fairness, spot them the OS, so that both teams have the same hardware budget.
I still don't believe that this is "fair." Benchmarks by nature try to convince you that their arbitrary imaginary scenario is fair. What I'm trying to propose is that context _matters_: one should start from a _real_ problem. For example, "the boss gave me a budget of $X, a set of requirements and told me that it's my ass on the line if the solution I concoct doesn't deliver" (or goes over budget - e.g., due to unforseen training needs).
Note that I'm not proposing that one or the other would win: I don't have enough personal information to make a reasoned hypothesis (I'd place a bet but not a bid). All I'm asking for is for some of the serious architects out there in Slashdot land to do the math and show how things would really stand up if we were playing for real money.
All that being said, I still get pumped when SGI posts a huge SPECfp-rate number (http://www.spec.org) - still my OS of choice for Big Problems(TM) $-).
And I hear my Octane is in the building now....
Actually, I've been waiting for the argument that seems to apply most directly: that benchmarks don't really represent the Real World(TM). For example, given the oft-cited media demographics of the "poor little pre-IPOs who just don't have enough money to buy M$'s latest work-in-progress," you'd think there would be some tracking of multi-function boxes.
;-).
For example, if you're using Linux as a major player at work, do you have a _pure_ web server box and a _pure_ file server box? Or do you have one or more boxes that really do lots of different stuff all the time?
This is a totally different measurement and gets directly to the heart of the "Bang for the Buck" argument: if I have $K dollars to spend on _one_ box to do everything, would I rather spend it on the best hardware (with cheap, powerful, easy to use) that I can get? Or am I going to have to spread it around on hardware+expensive OS?
At this point, you're talking real "business case" numbers that you can talk to your boss / capital approver and say "See, the path I recommend makes more sense in terms of total functionality, operator training, total cost of ownership, blah blah blah". (Anybody who's ever had to go through that drill knows what to put there - I just justified a $60K IRIX box for my desk... 8-).
So, the benchmarks that I'd really like to see are things like the following:
1. Given $5K, $10K or $20K to spend on a _single_ general purpose server machine, inclusive of OS and any serving technology, what's the best of class? Linux should win here by definition: cheap OS = big box, expensive OS = littler box.
2. Given the test platforms from #1 above, run through the Mindcrafty benchmarks again. At this point you're comparing dollars to dollars, so the results might be interesting this time. Not useful (steady state measurements don't indicate real life - does the Slashdot effect slowly ramp up or hit all at once...
3. Now show me something interesting. E.g., X web clients connected, file server traffic jumps up, then web clients drop off. You know, scientific method and all that?
At this point, I'd say the Mindcraft data is still far from being useful. Simply put, there's not enough of it.
Re: thanks - no sweat. It's what we professoresque types do.... ;-)
:-p
;-). Discussion of the ins and outs of grad school is a different thread, though....
Re: the non-compete, it's unlikely that it's enforceable / relevant outside of VA (maybe not even _in_ VA). The IP ownership definitely is, though, so don't play that game if you can avoid it. In the future, one of the best ways to defend yourself from this tactic (trying to steal your pre-job school-days ideas) is to publish _anywhere_. You don't have to put in every last detail if you think it's going to be marketable, but it proves prior art in one step ("as you can see, I was writing about this four years ago..."). Most CS departments have something like Technical Reports, if you don't want to try for something higher order (conference, journal or magazine) - a pretty easy place to make things a matter of record: here's what I've been working on, here's how it's going, here's where it's going, etc.
Final thought on the whole "those bastards are stealing my ideas" problem: do you really think it's likely that they'll be able to do a thing with them since you're leaving? I've had to leave code behind before and I get a big kick out of hearing from my spies about how the PHB just kind of stared at it: "Duh, what's this do?" "That's an rgbColor. You know, red, green, blue?" "Uh, yeah!" Yeah, best of luck with that, weenie boy!
Re: grad school - there's some really sweet graphics / VR / visualization schools out there these days: GA Tech is still #1 (unless I've been out of the loop long enough), Brown #2. Other examples: UNC Chapel Hill, UVA, Utah, Indiana Univ (go figure - they have a load of hardware, though, thanks to the efforts of former grad students
This is a bit spooky. I've lived this life: same place, different ... ;-).
:)
;-). Kind of like a professor, but I don't have to teach or ;-).
;-) \end{irony}
details. I'll spill the beans on how the story turns out: paying dues
early = victory in the long term (now define "victory"
More comments below:
> For the on-topic bit: I was suckered out to Virginia for $50k/year
> fresh out of college, but I've got lots of skill in realtime 3D
> which this company wanted, but they didn't know how to treat their
> employees and so I left after a few months.
I feel your pain. Note, however, that the little box on an HR drone's
checklist can now be checked: "Have you worked in your field outside
of college?" You also have a decent price point that you could use in
future negotiations, especially if you're planning on relocating: $50K
in Virginia dollars is worth less than $50K in Teeny Tiny New England
state's dollars. I've definitely done that math.
> For the off-topic bit: I have come to realize that the industry
> isn't for me. Academia is where I belong. I'm not a mercenary
> programmer. So of course, after a few months of living relatively
> large (figuring I'd be gainfully employed for a long time) I'm
> having my world kinda crash down around me, financially anyway. It
> doesn't help at all that I incurred some debt in moving out here
> which I, very stupidly, put off paying back. All in all, I'd have
> about broken even for the whole experience were it not for the
> various tech toys I suddenly found myself able to buy... Even though
> I rationally know that grad school is best for me, and emotionally
> know it as well, it just doesn't help to have all you mercenary
> types rubbing my nose in what kinds of salary I'm giving up.
This is where the popular culture seems to break down: the "industry
vs academia" dichotomy is really a spectrum. There's lots of jobs
that fall in between the two extremes: in my case, since last
graduation, I've been a post-doc, a "Scientist," a "Manager" and now
(just recently) a "Senior Scientist." Salary is more than double what
it was in VA, I'm off the evil "100% management" track (though I've
checked some of those boxes on those aforementioned HR sheets) and I'm
back into pretty serious R&D in my field (real-time 3D graphics - see,
spooky!
grade papers (hooray hooray!
> I've never been into computing and programming for the money, except
> for a brief period of time when I was graduating college and I got
> suckered into putting off my happiness for the promises of getting
> to keep doing the cool stuff while also making enough money to live
> very comfortably. Of course, those promises never panned out, and
> the company I got hired by turned out to be nothing more than a pair
> of two-bit swindlers doing whatever they could to control spineless
> employees who didn't know better and weren't at liberty to leave for
> a variety of reasons.
Again, I feel your pain. The suck factor tends to be pretty high
after college / university. The trick is to be constantly thinking
about where you're going, not just where you are, especially as the
suck factor tends towards maximum. Consider your next job interview:
1. You've worked in your field and you've set a salary point.
2. You've learned about how to do things _wrong_ in the business
place. If you can talk about this effectively (i.e., unemotionally)
with the next employer, you start sounding like a "seasoned veteran."
Obviously, it takes more than one job to get all that seasoning, but
still.
3. You're in a slick job market. \begin{irony} If you stay in
VA, there's only about five graphics people in the whole state and
they just seem to keep moving from job to job, doubling their salary.
Of course, the reason the supply is low is that the environment
stinks, but hey, money is god, right?
Likewise, I
seem to be the only graphics guy in this little Teeny Tiny State (see
above), so the last time I was on the market, I had a jumbo offer in a
week. Of course, I turned that one down because they tried to hit me
with a "don't work in your field for a year" non-compete, it wasn't
really in my field / area of interest, and the commute sucked!
Instead, I'm working directly in my field, doing the kinds of things
I'd want to be doing in the university (if I didn't have to teach /
get tenure), the company just bought me a brand-new great big machine
(costs _way_ more than double the price of my car $-) and I'm working
with people that don't suck. Sounds about right to me!
> I need to give some advice to academic types who might be reading
> this thread: which do you prefer, money or happiness?
Summary: it doesn't have to be an XOR. So sayeth the Academic Type.
> I used to process all the Security Clearances in Pacific Region in
> Canada, as well as declassify a lot of information in personnel
> files. And held a Secret clearance.
I'm on the opposite of the coin, re-activating my clearance again to
get back into classified projects (again). I've been dutifully
re-reading my briefing materials, especially about producing
classified documents, and have found them to be pretty up-front and
straightforward. Here's a summary (a techie summary, not an
"executive" summary) for those "not in the know":
1. Classified data = _real_ data. Approximations show up all over.
If you're including actual dates, frequencies, or multiple decimal
points, you're probably producing something that's >= Secret.
2. One classified number classifies that paragraph and, therefore, one
or more paragraphs, depending on your writing (notice, we're
talking paper here). Other pages may be unclassified (e.g., no
reeeally accurate numbers), but a document including the SuperDuper
Secret pages is SuperDuper Secret by the transitive property.
3. You are allowed (and encouraged) to produce unclassified documents
/ reports if at all possible. In the above example, take out the
really neat number and your paragraph, page and document level drop
to lowest possible (e.g., Unclassified).
4. Classification is not a toy (WillAffleck's example is depressing
but probably common). The documented government policy here is
"don't be a dick."
5. Classification exists for a reason. The best example I've ever
heard was when talking with mumble mumble about pre/during-Vietnam
War adventures and why he _still_ wouldn't talk about real details,
his reply was "They(TM) didn't know we could get a submarine in
there." I.e., if they didn't know we could get there, they don't
know _why_ We(TM) were there and what We(TM) were there for. Think
about it, when someone goes looking for something, it's because
someone told them "there's something to look for here." Those are
some of the "assets" that are protected by TopSecret++.
> Most, 95%, of the material classified as Secret is junk. The same
> holds for Confidential. I presume, based on inference, that probably
> 50% of Top Secret material is junk.
6. [Partial agreement] Confidential is silly: "potential risk to
national security"?! [I'm quoting poorly but I don't have the
definition in front of me.] The Jerry Springer show is a known
risk to national security (it decreases the average IQ of the
nation ==> bad ==> national security-- ), should it be classified
Secret?
Blah blah blah - this was a bit more than I intended but sometimes a
definition of terms is useful.
The punchline is: sometimes (often) declassification is a good thing;
just getting the paperwork processed decreases the entropy /
bureacracy of the universe. Sometimes, declassification is literally
not possible: in extreme cases, the "asset" in question is still in
"use" and we really don't want people to know how far the "asset" can
shoot, or where the "asset" lives / works / sends his/her children to
school, etc.
Finally, the easiest way to tell if somebody is working on something
classified is if every answer is a variant of "I don't remember." I
still play this game with mumble mumble above:
DB: "So where'd you go in that sub?"
MM: "Out to sea."
DB: "For how long?"
MM: "A while."
DB: "How long could you stay submerged?"
MM: "Long enough."
==
Doctor Bob
Clearance: I don't remember right now.
Project: Yes.