MicroTimes Interviews the Head Honchos of Silicon Graphics (1989)
MicroTimes Interviews the Head Honchos of Silicon Graphics (1989)
Ed McCracken and Jim Clark talk about their hardware and the future of 3D
Silicon Graphics is an interesting company and a pioneer when it comes to early 3D. Fellow Substacker
wrote an interesting post about the history of Silicon Graphics. Be sure to check it out.There were a couple of small errors and a couple of blank spaces below. They were in the original text.
from the March 6, 1989 issue of MicroTimes
3D Graphics: The Shape Of Things To Come
Silicon Graphics Prepares For The Volume Market Of The ’90s
On the screen, an aircraft rotates in black space, its sleek surface bathed in multicolored lights from multiple sources. It allows itself to be viewed from all angles, letting the viewer zoom in for a closer look or out for the grand view.
You're in a tank, threading your way down the corridors and around the corners of a maze—at a rate of speed and instability that sets the motion-sickness-prone to begging for mercy.
There’s a checkered room rotating around the screen. Inside the room is a little purple cube which purports to be made of Jello. As it bounces off the checkered walls, its shape changes to reflect impact and returns to normal as the cube becomes airborne.
“It’s really a mind-expander, isn’t it?”
—Silicon Graphics President/CEO Ed McCracken
By Mary Eisenhart
Watch out, Macintosh. Watch out, IBM. Watch out, Sun.
Silicon Graphics, whose high-end, RISC-based, 3D graphics workstations have for some years been a mainstay of niche markets from video to mechanical design to aerospace, is taking aim at the under-$20,000 workstation market. The key element in its strategy: the Personal Iris, which starts just under $16,000. It runs the same software as SG’s high-end, multiple-processor systems and, like its more sophisticated and expensive brethren, manipulates three-dimensional graphics in real time.
It’s hard to exaggerate the sheer emotional impact of your first encounter with the Personal Iris. Especially if you’ve been in the habit of going to the mall while your Mac II struggles with fractals, or wishing that wire-frame model on your 386 bore some resemblance to anything in the real world.
The Personal Iris, which runs Irix, Silicon Graphics’ licensed version of UNIX, includes some strategic features not usually seen on UNIX systems. There’s a comfortable window-based interface. The system has been designed so that a naive user can have it out of the box, set up, and doing real work in 20 minutes. The RAM upgrades (standard configuration of the Personal Iris includes 8 Mb of RAM and 8-bit color) are user-installable. If something goes wrong with the machine, all its electronics are contained in an easily-removable: box; remove it, FedEX it to Silicon Graphics, and a loaner box is on its way. Finally, in contrast to most UNIX machines, which arrive with an array of manuals rivaling the Library of Congress, the Personal Iris comes with a small book containing everything a user needs to know to operate and administer the system. (A full set of documentation is available for programmers.)
Currently a quarter-million-dollar company, Silicon Graphics is aggressively campaigning to open new markets by reducing the cost of its technology. At the high and the low end of the product lines, the company seems obsessed with delivering ever-higher performance at an ever-lower price, and becoming a key player in the evolution of the computer industry. And if, as is the plan, small businesses and individual designers come to demand technology you used to have to work for NASA to enjoy, we'll be hearing a lot more from this other, quieter workstation company from Mountain View.
We've been watching the progress of the Personal Iris for some months now (it was unveiled in October, and Silicon Graphics recently shipped its thousandth unit). Recently we visited the company’s Mountain View headquarters to discuss its products and strategy with President/CEO Ed McCracken and Chairman Jim Clark.
We asked McCracken about the Iris line, what its current applications were, and what uses he envisions for the future.
What have been your markets until now? Who’s been: using your machines for what?
Ed McCracken: One of the nice things about three-dimensional graphics is the pervasiveness of use across a broad number of industries and applications. The biggest applications for our machines have been the automotive and aerospace market for different types of mechanical design and mechanical engineering. That’s representative of 50 percent of our business. It’s not the markets that we penetrated first, but it’s the market that is the biggest, and we think it will continue to grow with us and be 50 percent of our business when we reach 500 million and when we reach a billion dollars. There’s a lot of robustness in that one marketplace.
The average mechanical engineer today either uses a terminal to an IBM mainframe, or they use a two-dimensional workstation. And they’re dealing with three-dimensional structures, so it’s awkward. Our systems can make a huge difference in the ease of use of the mechanical designer and how long it takes for them to get the job done. As a result, we think it makes a big difference in the competitiveness of corporations that commit to a three-dimensional approach.
That’s the biggest. The first market for us was animation and creative graphics, and we feel that we have about 75 percent of the workstation portion of that market. It’s about 10-12 percent of our business. Our current run rate is around $260-270 million total, so 10 percent of that would be maybe $30 million a year, which gives us most of the market. It’s not a huge market; it will continue to grow some, but the nice thing about that market is that those people force you to have the best possible, most realistic image on the screen, and move it around in real time. That forces us to be on the cutting edge of new graphics architectures and new approaches to graphics. So then we take those capabilities and essentially do a technology transfer of those capabilities into the other markets, which are a lot slower to decide they need it, and that puts us way ahead in the other markets.
So that’s another application. The third. is gaming and simulation, where people use our product to simulate environments. The military was the first user here, where you use it to simulate a view out of a pilot’s window, or those types of things, or to do mission planning, or to design a new cockpit of an airplane. But that’s moved into simulations of manufacturing floors, and simulations of robots, and it’ll eventually move into things like driver training and games, a lot of games that have a learning objective associated with them. That’s about 25 percent of the business.
Then all the rest of our business, essentially, is in science. Visualization of chemistry-physics types of things. That’s a very robust market for us, and it’s an important market because it’s leading edge.
We made a major transition in the company over the last year in that our products used to be ahead of everybody else in graphics, but the computing was, I guess you could say, marginal for most problems.
Because it was optimized for graphics?
It was optimized for graphics, and we just didn’t pay a lot of attention to the computing, and then it became evident to us that the thing that was holding the graphics back in terms of performing was how fast we could feed the graphics data. So we started saying “Well, maybe we should increase the speed of our computing system.” And so we kind of put our heads down and worked real hard to do that over the last two or three years—and then recently we kind of raised our head up and looked around and found out that the computer we had invented in the process was faster than anything else on the market in anything near the price range. It’s that computing speed that’s causing the physicists and the chemists and people like that to really beat a path to our door right now; they need the graphics, and the visualization is great, but the fact that it also is the fastest computer you can buy for anywhere near this price...
This is at the high end or the low end?
This is at the high end.
So those are the primary applications. There are a lot of new applications that are kind of innovative—they don’t add up to a significant percentage of our business yet; they might in the mid 1990s. Applications like using our product at the point of sale, where maybe eventually you’re going to buy a car and you'll sit at one of our systems. You'll say, “I’m interested in this model,” and it pops up on the screen; you drive around a bit on the screen, and then you say, “These are my color options,” so it colors it the way you wanted it, and “These are the other options I want,” the texture of the seat covers, and so on. You choose your car exactly and it prints out your invoice.
Or you go in to buy some kitchen cabinets, so the whole catalog of kitchen cabinets is available to bring up on the screen. And then you define your area that you need to put the cabinets into, and it pulls them up, and when you get it all right, it looks the way you wanted it, then you print your invoice for kitchen cabinets. That’s another one.
There’s another new one in color printing, in terms of using our systems, which are great color systems, to help people electronically with each color print. And then the architectural market is going to get interested in our machines shortly, and I think there’1] be markets that expand out of that into interior design, and landscape architecture and all kinds of unique, lower-cost kinds of markets.
I read that someone was using it for fashion and textile design.
Levi Strauss in San Francisco. They’re really saving a lot of money doing it, because they can take a picture of a model wearing a sport coat, and then they can say, “Well, I wonder what that would look like in a plaid.” They can take a plaid material and put it over her, and they’ve figured it out so the plaid material will fit into the creases, in the armpits and so on, and look real. It’s a pretty wonderful thing. And then if they decide they like it, they can go straight to the catalog before they start production, so they don’t have to do a sample.
One of the things I keep hearing is that 3D is useful because it saves a whole lot of expense on prototyping.
Yes. Absolutely. I was just talking to some of our people about a major Japanese company which has a large number of our systems that they use for power plant design. They have about 100 of our systems, I think, and before this they had to make these complex models of a power plant that took up an area almost the size of a basketball court—plastic models that were that big. They’ve been able to eliminate that whole process and go directly from the models on the screen to the construction site. They can give sales presentations to the customers and so on, all on screen, and never build the models. And be able to do all the checking to make sure the pipes aren’t crossed, and that you have all the room you need to do the pipefitting and so on, and that you don’t wait till too late in the project to take in this pipe that has these curves in it because you can’t get it through the other pipes to get it to the right place.
It’s really neat stuff. And they’re being much more competitive in their markets as a result, to the point that they’re almost putting some of their competitors out of business.
Why don’t their competitors get with the program?
It takes time. There is a learning curve involved.
It turns out that in using our product—using three-dimensional databases and three-dimensional software running on three-dimensional hardware because all three are important—the first phase is, you get on the learning curve. First you buy the systems and you experiment with them a little bit.
And then you move into phase two. In phase two, where most of our customers are, if the product design process is that you do some conceptual design, and then you do some simulation and some analysis and some plant floor programming, you use our systems in any one or all of those processes, and it makes each of those processes more efficient.
Phase three, which this Japanese company is into now, is when you say “Aha! I’ve got the database captured electronically, I can pass it back. I don’t have to do serial design. I don’t have to finish this design before these people get started. I don’t have to pass it down the serial process; I can build a team made up of all of these departments, and! can put them all together and they can pass data back and forth electronically, and can turn the whole design process into a parallel process rather than a serial process. And maybe I can cut my design time by a factor of four. And as a result of that I can be more competitive in world markets.”
That’s where you finally get to in the learning curve, but if your competitor’s doing it all manually and on two-dimensional workstations, and you're doing it in phase three, then they’ve got to go through phases one and two in order to get to three, and so you're way ahead of the game.
So there is a very important time issue in, competing with companies using three-dimensional computers. The sooner you get involved the better off you are in our competitive world.
This may be a little out there at this point, but it seems to me that one of the natural applications of this technology is hypermedia. The Iris looks like the closest thing to a Xanadu machine I’ve seen yet.
It is. The hardware capability is going to be there, there’s no question about that. It’s going to take some software work. Each application, has a slightly different hardware/software cycle. The advantages. of three-dimensional activities in mechanical design were so important ten years ago that the major software vendors invented three-dimensional software and three-dimensional databases, which they’ve been running on two-dimensional machines for the last ten years. And now that they’ve got three-dimensional hardware, it’s phenomenal.
In other markets, the three-dimensional hardware is there and the software is just starting to be developed now.
But the Personal Iris at its price point is going to allow a lot of very creative people to do creative software. Some of them are here at our company, but most of them are out there in the world. Hardware capabilities aren’t going to be that difficult to put together.
What about virtual environments? They’ve got a helmet at NASA that puts you in a virtual room, where you can walk around and pick up objects. Do you see that type of application opening up as a result of these machines?
That’s possible. The helmets at NASA are driven by our systems. They’re hidden off to the side somewhere, because you can’t put the system on your head, but the helmets use little pads that are in front of your eye and display the images on the pads. They were black-and-white the last time I was there; maybe they’re color by now. Our systems drive those.
There are applications where I think our systems will be used with that kind of user interface rather than a screen. But who knows, in the 1990s, what will happen there? It’s certainly one of the things we’re watching; our systems are being used as experimental devices.
I have trouble myself believing that the average architect will come to work and put a helmet on their head and design their house. It’s possible, who knows? There’s no reason why it’s not technically possible.
You do see that screen and want to jump in.
Right. You really create an artificial environment of these ideas, so the ideas you have about this city block you're designing become your reality while you’re designing it. You can see how it would really unleash some nice creativity.
I can see game developers flipping over this machine if you can get the system price down.
Yes, and especially with fiber optic technologies as they’re developing, you could have group games that could be really exciting, where you’d not only exist in the space, but multiple people exist in the space. Which would be kind of fun.
The Personal Iris is really our fourth-generation product. We’ve been working on three-dimensional graphics in computing for about that long, and it’s so different from our first- and second-generation machines. Our first- and second-generation machines were primarily used in wire-frame mode. Everything was vectors, and we quoted vector rates and line-drawing rates and those types of things. And we’ve gone so far beyond that now, because we think realism is important.
There are a lot of other companies in the industry that I think will be introducing some “three-dimensional” products this year, and they will primarily quote vector rates. I think the market’s going to be a little confused sometimes about what they can do versus what the Personal Iris can do. We’re trying to simulate reality on these systems. We’re trying to make the image real—you can’t do that with stick figures, except in electrical engineering, where if you're dealing with schematics or something, you might be able to get by. But for the applications we’re interested in, you need lighted, shaded surfaces that can operate in motion in real time.
Silicon Graphics executives are fond of saying “If it don’t move, it ain’t 3D.” Could you elaborate on that?
Earlier I talked about a lot of the mechanical CAD companies having developed 3D software during the last ten years: because they saw the need for a three-dimensional database and for the three-dimensional software to operate on the database. So they would sometimes call their applications 3D. And they are 3D, except when you went to use them on any workstations that existed, they could only draw one image every second or every two seconds. Or even every ten seconds. And as a result you couldn’t move it on the screen.
If you can’t move the image on your screen, you can’t get a feeling for the third dimension. You may be able to take a three-dimensional database and bring it out on the screen, but once it’s there it’s just a two-dimensional flat picture. And in order to really understand the depth of the image you've got to repaint it at 20-30 times a second. When you see a movie, you see 24 frames a second, and 24 frames a second is about the right speed to view an image at.
That’s why the graphics for three-dimensional computing is so much more complex than two-dimensional workstations—you have to repaint the screen 20-30 times a second. So there’s a lot of controversy about what really is 3D, and we just make it easy: if it doesn’t move on the screen, it isn’t 3D.
What's the difference in the speed of image movement on the Personal Iris compared to the speed of your high-end systems?
It isn’t so much the difference of speed, it’s how complex the object can be and still move at 20-30 frames a second. I think there’s probably around the order of five to ten times difference in the complexity of the object that you can show on the screen, moving in real time. So if you can show a house on the Personal Iris and move it, on the higher system you can probably show the house, and you can probably have windows on the house, and you might be able to see through the windows and have the windows be translucent. Maybe you could have a tree, and maybe you could have a few other things, and still be able to move around it. On the Personal Iris, maybe it’s just a house. That’s it. The windows aren’t translucent.
Light sources are pretty much the same?
The same algorithms are used, it’s just that we use so many more processors on our high-end systems that we can do things much faster. But one of the nice things about our product line is we use the same algorithms top to bottom, which means that the same graphics software can run on all of our systems.
Which is a great thing for the software companies, because they don’t have to have different versions of our systems, like they do on most of our competitors’ machines. They can use the same software across the entire line, and it’s all binary compatible from both a computing and graphics standpoint.
It truly does change everybody’s experience of ever using a computer in the history of computing. History of computing: 2000 BC till today.
In your mind, you create images and thoughts and feelings, most of which have depth, and historically you’ve had to squeeze it and change it and linear-ize it—generally move it from your right brain to your left brain, get it out into symbols the computer understands, and then the computer chunks away on it and gives it back, in symbols the computer understands, and then you take it from your left brain and try to feed it back to your right brain. And of course by now your right brain’s asleep.
As you know, creativity has momentum, and when you can keep your right brain of your mind enraptured by something and going on it...
Once you find that groove you want to stay with it.
You get the groove, you get the high, and you kind of go for it, and you make these trememdous (sic) leaps. It’s our belief that you’ll be able to do that with this class of system.
Of the customers who are considering a PC or Mac II workstation, or maybe a low-end Sun/4, who should look at the Personal Iris too?
We're repositioning our company. Nine months ago we wouldn’t have been on the map. We wouldn’t have been considered at all. Today we are being considered in a lot of applications, and I think we will be considered even more in the future. A lot of it just depends on whether you’ve heard of Silicon Graphics or not, and that doesn’t vary all that much by application—it’s just that we’re so new in this marketplace.
Our prices are not all that different. The price of our system and the price of a Sun/4 is about the same. And if you load up a Mac, as a lot of people do at the high-end applications, then it gets to be about the same.
So the person that will tend to use...(sic)who’s got one of three requirements. One is that they know they have significant computing needs, that their application burns computer cycles. If so, there’s nothing faster anywhere near the price range than our system. There’s nothing faster at two or three times the price than our system, just in terms of cranking away at the computing behind the applicaton.
There isn’t a better full-color system on the market, so if you really want 24 bits of color and you want that handled really nicely, our system is great at that.
And the third reason is if, in what you’re doing, you want to deal with anything that has a third dimension, anything that has depth. Our system handles the shallow stuff fine, and a lot of the full-color thing is just two-dimensional stuff, and if you need full color our system is great. But if you want to do anything with depth our system is the obvious choice.
And of course it has a lot more screen resolution than a Mac, also. Even the Macs with all the add-on boards get up to maybe 1024 by 7-something, and ours are 1280 by 1024.
In order to do all this graphics handling and fast computing, you need a lot of RAM.
Yes, we do. Video RAM, mostly.
That’s been something of a problem lately—has this affected you significantly?
Well, yes, it has. When we set out to design the Personal Iris two years ago, we expected it to come in closer to $10,000 than $15,000. We thought by now we'd have been able to price it down in the $10-$11,000 price range. We had to introduce it at $15,000. The big difference is RAM price.
We use two different types of RAM in the system, dynamic RAM, which is like everybody else’s, for the main computer; and then we use video RAM, which stores the information right before it goes onto the screen. So it has the information about color, and what’s in front and what’s behind, and those types of things. Those have been in short supply. We’ve been able to get enough to meet our needs, but the prices are certainly a lot higher than we expected them to be two years ago.
The Personal Iris seems a lot more accessible to relatively naive users than most workstations.
We’ve gotten tremendous feedback. We shipped 1,000 systems last quarter in a 90-day period of time, and we’ve gotten tremendous feedback from our customers that they were able to take the systems out of the box, put them together, get them running, in less time than they could a PC or a Mac, and a lot less time than a Sun product. The system is much more approachable, and that was our objective. We spent an awful lot of time with that. We had a lot of videotaped focus group work, where we brought people in and had them do all those things and be videotaped, and we made signficant changes to the product, in the packaging, and software, and the boxes they came in, and the manuals, because of that videotape work we did. It was really effective.
You have user-installable RAM upgrades.
That’s right, we do. And we’re getting very positive feedback on that.
Nobody’s frying their motherboards?
Nobody’s frying their motherboards. It’s even better than we thought. It’s a little surprising to us, because we were doing it as kind of an experiment. This is working quite well—we did some focus group work on that too, and as a result we changed the sockets a little bit.
Where do you plan to go from here with the Personal end of the product line?
Well, we’re not pleased with our current prices. We think the performance is nice, but we’re not pleased with the performance either. So the $15,000 barrier that we’re currently at is going to break, and we'll get the prices down to 12 or 10 or nine or eight [thousand]; we’ll take it down as far as we can. We don’t think there’s ‘anything inherently more expensive about a RISC machine than an Intel- or a Motorola-based machine.
We think that in the 1990s there’s no reason that you should pay any more for our low-end machines than anybody’s PC—powerful PCs, of course—and the performance should be really dramatic.
The other thing is, even our systems don’t have the kind of realism that customers really want. If an architect is designing a house, they want the house to look real. It looks pretty real on our system, but there’s no trees in the yard in front of it, and the walkway and the street and the texture on the side and so forth—if you put all those things in, and you make the house really complicated, you can’t really rotate it in real time, like you’d really like. There’s an awful lot of things from a graphics standpoint that we plan to make better, both in terms of more performance and more features.
And then the third thing is just software. We plan to put a lot more emphasis on software. We have an internal objective to go from 400 packages, which we have today, to 4,000 packages in three years. So that’s taking a lot of our energy and a lot of our time.
Maybe the last thing is we’re readying distribution channels. We’re in the process right now of signing up dealer/resellers—I think what you’d call VARs. They are geographical in nature, they’re not nationwide, and some of them have storefronts, some of them don’t. We’re following our software partners into them. So we’re working with CADKEY and Sigma Designs—CADKEY is in mechanical design, Sigma is architecture—and we're going out to their resellers. Our objective is to have 50 of them signed up by June, which is a good start, we think. It’s a test program between now and June, and then we'll expand it after that.
So that’s a new thing for us as well, and of course that would become more and more important to us as our prices come down.
Are there any particular types of software applications you’re interested in attracting?
Well, yes, obviously, there are. Anything in architecture, are very important to us, but beyond that we specialize in chemistry, we specialize in anything in the publishing area, anything, certainly, in the gaming and simulation environment would be good.
But more generally, we believe that the really significant software that will run on our system, and other people’s three-dimensional systems in 1995, hasn’t yet been written. So we’re really out trying to convince people to write three-dimensional software, because it’s a lot more fun to use, and it’s fun to write, and it’s going to be the volume market of the 1990s.
What are the demands that 3D imposes?
I think the opportunity that it creates is that you can assume you can move whatever you have on the screen. It changes the nature of the way you write the software, because you assume motion. So you're creating a movie at the time and so you write it a little bit differently.
Beyond that, it’s not that different from writing any other kind of software. You have to pay attention to the object itself, the environment in which the object exists, how the object is lighted, what kind of textures it has. Those types of things. Which a lot of programmers aren’t used to yet—on our system we have something we call our three-dimensional tutorial, where you can try out how to program to change the texture of an object, and how to program to change the lighting of it, and whether it’s a pinpoint light source or a more diffuse light source. We build a lot of those types of things into our systems.
For somebody who’s essentially a computer illiterate, how long would it take them to be doing useful work on these machines?
It depends. If you’re a computer illiterate, it means you're not going to use the system unless an applications package exists that you need to use. So it depends on the applications package.
If you’re a chemist and you're a computer illiterate, and you're at Genentech and you're going to be designing molecules of certain types, then you'll buy software from one of about four sources, all of which run on our machines, and maybe in about a month you'll get pretty comfortable with it.
Or if you’re an architect and you bought the Sigma Designs software, you'll probably be comfortable in a few hours. It varies by market. If you're a film animator, it may take a little longer, because the software’s pretty complicated, and that kind of software tends to assume that you’re a genius, because the film animators tend to be kind of creative geniuses. It varies—each software package has a little different learning curve.
We're shipping something we call Quick Model and Quick Paint with our system, which gives the computer illiterate a chance to bring the system up and build a three-dimensional model and animate it without any knowledge of a computer at all. He should be able to do that in the first hour—which gives you a chance to feel successful, which is really nice. You don’t have to do any programming to do that, you just have to choose options on a set of menus and icons.
And then we have WorkSpace, the icon-based user interface on the Personal Iris. It makes UNIX a lot easier to work with. Initially it’s on the Personal Iris, and eventually the whole product line.
We have a lot of very nice icon-based UNIX administrative tools on it. When OSF or the standards organizations agree on exactly what the user interface that goes on top of UNIX will be, then we'll overlay that on top of what we have. But this will exist until that time, or maybe even beyond that time, as our primary user interface.
The Big Picture
Silicon Graphics Chairman Jim Clark On The Workstation Marketplace, Standards, And 3D
What needs was Silicon Graphics intended to address?
Fundamentally we were driven by the lack of real usable graphics systems of the type that I had, always been exposed to—real-time three-dimensional systems, basically, doing things and being able to connect with real-world physical problems in a: very interactive way. So graphics has always been the drive of this company, in particular 3D graphics, but I would say that over time that'll probably broaden quite a lot.
I like to think of the company as sort of the graphics computer company—not the computer graphics company. We sometimes focus a lot on 3D because it turns out a very significant part of the world is three-dimensional, and it’s the part that really hasn’t gotten developed.
Designing three-dimensional things has always been a pencil-and-paper type of discipline. The currency of trade of a lot of three-dimensional engineering has been the drawing. But over time, with more and more computers at your disposal, the drawing on the piece of paper will become less important and the media of exchange will be the computer. When you’re dealing with a three-dimensional object, you want to be able to exchange the three-dimensional object, which you obviously can’t do, but you want to be able to convey to a manufacturer on a floor, for example, “Here’s what the part has to look like.” And to the engineer who’s doing the engineering, you have to be able to convey “Here’s what the stresses mapped onto this object look like if this object is put under certain kinds of loads.” So it’s a kind of soup-to-nuts desire, is what the objective of the company is now, and it focuses largely in the CAD/CAM/CAE area, but it’s going to evolve into a lot of other areas, such as chemistry and architecture and all of these disciplines that are three-dimensional.
The roots of the company really are driven a lot by graphics, but you can’t have graphics in isolation. The whole workstation industry came out of the merging of computer graphics and computing into one box. To some degree the same is also true of the personal computer industry. It is the principal interface to a computer, your visual system, so having a very high quality graphics system is really important. We’ve been the leader in that area in the three-dimensional area, and we intend to remain that way.
Who are your competitors in 3D?
Competitors tend to come in categories related to a particular application, or a particular installed base that they’ve already developed because they were here before with computer systems, or they had a two-dimensional product that they’ve gotten into accounts with and they’ve grown rapidly. So you can look at it on the one hand and say there are no competitors, but then realistically every computer company is a competitor, and every workstation company is a competitor.
On a one-on-one basis, if the requirements are three-dimensional we tend to have no competition. I suppose in that dimension the company that has surfaced most has been Hewlett-Packard. Hewlett-Packard doesn’t seem to have a competitive product line relative to us. They just basically don’t. At a given price point, we’ve got a much better product in the 3D area.
Who needs 3D?
Who doesn’t?
The world’s going to be a very different place in another 10, 15 years—the way computing is done, the way engineering is done. And all engineering that entails designing or understanding some kind of three-dimensional phenomenon will evolve to the point where it has to have and will use a computer, to both visualize and do the actual design in a more upfront three-dimensional manner. Today, the starting point for design, for the most part, is a drawing. Even when you go to a 2D workstation, it’s still the starting point.
There are lots of disciplines and lots of areas of applications that aren’t three-dimensional—desktop publishing is not three-dimensional. Spreadsheets. I have a personal aversion to trying to find a way to present data that’s not intrinsically three-dimensional in a three-dimensional form—for example, a three-dimensional view of financial data, or three-dimensional bar charts.
I don’t think that serves a purpose, and it certainly isn’t what drives us when we think about this. It’s more—what is the discipline innately dealing with? In architecture it’s obvious, right—it’s a building. In designing industrial products—chairs, coffee pots, packaging for any kind of product—it’s a three-dimensional thing. Automobile design, styling, down to the levels of mechanical engineering—that’s all three-dimensional. Aerospace industries—it’s all three-dimensional. Chemistry—we've been making stick figures in chemistry. Remember when you took your college chemistry class and there was always this little stick figure? It’s a geometric process.
Medicine—a lot of medicine, a lot of reconstructive surgery modeling being used, and understanding three-dimensional images, CAT scans and NMR images. Those are really three-dimensional images, as opposed to geometry. You infer the geometry from the image—X-ray is a two-dimensional image from which one infers things about the bones and internal structures of the body, and likewise NMR or CAT scan type of technology is a two-dimensional imaging technology. Nonetheless, people derive from those images 3D shapes which are then represented as geometry. For example, some of the reconstructive surgery and hip joint manufacturing—finding the precise angle for the hip joint in order for it to fit a particular person, sort of customized hip joints as opposed to standard-part hip joints. I know there’s some people back at Cornell using it for that.
You design the part onscreen and then it’s manufactured to the computer’s specs?
Yes. The point is there’s an X-ray, or maybe even an NMR scan, of this person’s hip, and the idea is to get the precise dimensions of the ball, and the precise angle that the ball needs to go away from the socket, or that the piece that connects to the leg has to go away from the socket in order to be a proper fit for that particular individual. All of us are slightly different in our geometry, so it’s really important for that.
So my answer to the question is that there are certainly some disciplines that aren’t 3D, but by and large engineering, as I see it, and science are three-dimensional.
How does a software package like AutoCAD in the micro world fit into what you’re talking about?
It’s a drafting package. There’s a little company called ParaComp that has a couple of products I would consider real 3D products. One is Swivel 3D and another is Model Shop. Swivel 3D is a little animation/conceptual design modeling package for creating simple geometric models for animation purposes. Model Shop is an AEC (Architecture/Engineering/Construction) package that’s a conceptual design package. It allows a conceptual three-dimensional model to be constructed right there on the scene of your new architecture. The thing that you have when you're finished is this three-dimensional model that you can get renderings of, see from different points of view, take walk-throughs of—it’s for the up-front conceptualizing phase.
After that phase you need to go into the drawing phase. So it’s really something that would work as an adjunct to AutoCAD. You do need drawings in the architecture business. That’ll change. Specific drawings as we use them today will change significantly over the next years, but for today it’s still the currency, as I said before. Exchange between a contractor and an architect or whomever.
So AutoCAD is focused on today’s need to have drawings. It’s been pretty successful there. But they aren’t a 3D product. They’ve tried to add some 3D capabilities, but after the fact is a little more difficult.
What are the capabilities of your various machines?
Personal Iris is the highest performing workstation available with 3D graphics in the under-$20,000 price range. It’s 10 times the old VAX 11/780 in basic compute power, and it’s got computer graphics power that’s about 30 times faster than what we first introduced four years ago. So it’s a pretty powerful little machine, the one that a lot of people are most excited about.
Well, not that many people are in a position to go out and drop $80,000 on a high-end system.
Exactly. Well, that was the reason for doing it. We would have had a low-end machine a long time ago, but we chose not to, because we wanted to make sure that we continued to make 3D machines. The technology of integrated circuitry, and memory, and so forth, is now at the point where you can produce a 3D machine at a low enough price. And we're aggressively pursuing making them available at lower prices yet,
That’s a real important thing for us, to continue to make 3D machines, and compatible. Right now Silicon Graphics is the only company that offers a totally compatible product line. Every other computer company, every other workstation company, has two or three different CPUs in their product family, including Sun—at least three, the 386, the 68000, and SPARC. Whereas ours is Mips CPU top to bottom, and we're the only company that has object code compatibility.
So there’s a lot of similarities, but object code compatibility means you can take binary file written on one machine and run it on another machine. So a file or a source program that’s developed for the Personal Iris can be compiled and put in a runnable format, and you can take that format and immediately run it on the high-end machine without changing it.
So you can preserve your software investment.
Exactly. Software developers develop code for one machine and they can run it on all of them. It’s a story that DEC used to have, in the compute side only, not the graphics. DEC used to have this story for computing on the VAX/VMS line. And right now we're the only company that has that. It’s turning into a pretty powerful marketing tool for us, to convince people to develop more software, and to encourage people to buy from us in a given
competitive situation with another company, which might have a product that’s moderately competitive at that particular point, but they don’t have the rest of the product family.
A good example there I think is probably some of the high end competitors, and there’s a bevy of them, that are fairly noisy and strident but haven’t been very successful selling against us, not only because we seem to have a faster product at the high end than they do—I think that’s probably a surprise to them—but also because we have the product family and they don’t. So it’s a powerful set of processors; it spans the range from $20,000 up to about $150,000.
You develop your own system software, but not application software, is that correct?
Right, that’s correct.
So how do you work with application developers?
Very carefully. (laughs) We’re partners. Our livelihood depends on their getting software on our platform. So we work with them through a program called the Geometry Partners Program, which has a variety of different points, one of which is putting aside money to help them do co-advertising with us, as credits, in a sense. Providing them machines at discounted prices so they can get software up. A variety of things like that.
All your machines are RISC machines at this point?
Yes. We’re also probably the only company with that distinction. One hundred per cent (sic) of our sales are on RISC machines.
Probably the most well-known RISC architecture is SPARC from Sun. What led you to go with the Mips processor rather than SPARC?
The Mips design is better. That’s the reason we went with it.
Better in terms of what?
Faster. For a given clock rate, you get more work done. So if you fabricate both chips on the same technology, and they can both run with the same clock rate, one-does more work. The Mips chip does more work, that’s a fundamental difference. Running UNIX, running standard applications, it delivers more work per unit time than does the SPARC chip using the exact same clock rate, by somewhere in the range of 20-30 percent. Significant enough that, for example, at the same clock rate we get 10 MIPS and the SPARC architecture tends to get around seven or eight MIPS. So it’s not insignificant.
Aside from that—that’s just sort of the bottom line—the other thing is that Sun has throughout its history been fairly—Sun is sort of after all markets. It sort of thinks that it can take over all of technical computing. And at the time it was being offered and we were making the decision, we could get the Mips chip and we couldn’t get the SPARC chip. Mips was available beforehand. Even if they had been available at the same time, there was the clock rate performance difference, but even if it hadn’t been for those two things, we probably would’ve looked for an alternative just because it was coming from Sun.
On three or four different occasions in the last four years they’ve said “This is the board that we’re going to stick into Suns and sort of negate the need for Silicon Graphics to exist.” They’ve been pretty aggressive about trying to come into the 3D market, but they just haven’t been able to do it.
Connectivity and multivendor environments are a big issue these days. How important is connectivity in the environments you’re operating in?
It’s essential. But every company that’s trying to be a major player in big corporate America and corporations throughout the world has to supply all that. It’s a given. What you’ve got to do is add the real-time 3D and the software to make a difference so that you can fit in those environments. You have to be able to. We’re a licensee of a lot of these different technologies—for networking NFS, for example. We use a lot of that kind of technology.
Connectivity is crucial. At McDonnell-Douglas, connectivity to IBM is important. At Boeing, connectivity to the Cray environment is important. At NASA, connectivity to the Cray is important. Connectivity to the DEC world has always been an important thing, because they were such a dominant technology computer company. But you’ve just got to do a little bit more.
We spend our energy working to make our graphics better, let’s say, and graphics architectures faster, whereas Sun might spend more energy developing a CPU, the SPARC architecture, because they felt that was necessary for some reason. We try to license compilers and get the kind of stuff that a lot of other people invested a lot of energy to make, rather than developing them ourselves.
Concentrate on what you do best already.
Exactly. Where our unique value is, concentrate there to continue to be uniquely adding value in a way no one else can.
Do you foresee that 3D is going to be something that people come to demand in general computing as well as technical?
Oh yeah. No question about it. It’ll be on all machines—but who will have the best? That’s the question. And it’s our intent to be the company that comes rolling off your tongue when you ask that question.
Think about the computer industry in the last 20 years. The companies that sort of rose out of the ashes there on the basis of compute technology—DEC, Data General, Prime, CDC, IBM, Unisys, Burroughs, and companies throughout time. They dominated on the basis of some unique architectural or technical edge—like DEC, for example, just built a better machine than some of their competitors. Data General just kind of lost it, lost whatever their formula was for doing good machines. And then others benefited or grew, not so much on the basis of technology, but just by going after market niches, big commercial market niches that IBM wasn’t successfully addressing, or was ignoring.
We’re in a technical business. We want to remain competitive on the computing side of things, but absolute leaders on the graphics side of things. For the last while we’ve had the fastest workstation, the fastest computer system on the market for less than $20,000. A lot of people don’t associate the best computing with Silicon Graphics, but in fact at the high end we’re offering the best computing, the most performance in technical computing you can get for $100,000.
If something's going to bounce squishy jello cubes for $16,000, it’ll probably do your spreadsheets pretty fast too.
That’s right. Our compute capability is extremely good today, and we hope to keep it that way, but we’ll keep our leadership in the market by having the best graphics. I think there’s still big opportunities to grow and always be the best. We’re known right now for that.
Your operating system Irix is UNIX System V.3?
System V with a whole variety of extensions, including virtually all of the Berkeley extensions.
Are you going to go to V.4 when it comes out?
We’ll go to the best one. V.4 is the most likely, I would say. AT&T, as far as I’m concerned, is still the source for UNIX. I tend to agree with what Scott [McNealy] said in his interview [“The World Ac- (sic)
[CAFP editor’s note: This is a reference to an interview that MicroTimes had with Scott McNealy in December 1988. That will be next month’s interview.]
Who needs it?
Well, you know, I can understand. The bottom line, the reason that whole thing got polarized like it did, is because it appeared to the rest of the world that Sun was trying to pull a fast one with AT&T.
If, in fact, UNIX development fell under Sun’s control, which is the way it seemed at the time that the announcement of the alliance with AT&T and Mr. Cassoni’s joining Sun’s board of directors and so forth—it seemed that Sun was essentially going to get the control of UNIX, and that should worry the rest of the world. Contrary to whatever they say—they’ve got great marketing statements, Sun, but in fact they’re out to take over the entire computing world. And if they see UNIX as a vehicle for doing that, they’re going to do it. So there’s nothing benevolent about their behavior.
So that caused this other reaction. They’re the ones that caused OSF to come into being, is what I’m getting at. And if in fact Sun and AT&T did stay in this tight alliance, I think you would find real merit in what OSF is doing. They would have to find merit, because it couldn’t be at the disposal of Sun.
As long as AT&T is kind of this harmless vendor of UNIX, harmless to all computer companies, so long as they can retain that somewhat unbiased posture, it will continue to be a standard based on their UNIX. And there will be no reason for OSF to exist. That’s what’s ironic about all that—Sun complains about it, they poke holes at it and shoot at it and so forth, but it will absolutely be successful if they end up going off and causing a polarization.
So they’re really in control of it—all they have to do is relax it. Well, AT&T is more in control than Sun is. I think their whole thing was considered—from my perspective, at least, it was considered a mistake once Cassoni left and went back to Olivetti. It’s my perception that everyone was kind of looking for a way to gracefully get out of the deal with Sun in some fashion.
If were in AT&T’s shoes, I sure wouldn’t want to put all my eggs in Sun’s basket. Sun is a very successful and good company, but there are lots of other companies who are going to use UNIX. Unless they have another agenda. They may have another agenda, like acquiring Sun, or having Sun be their principal marketing arm for computing.
How much of Sun’s talk about building open, cloneable machines do you honestly believe is real, and how much do you think is a marketing position? If there truly were a machine out there offered by the Japanese that had access to every single piece of what’s available from Sun—all of the window management tools, the networking tools, the program development tools—and were truly cloneable, they would shake in their boots. It’s total garbage, what they put out in that sense, I think. I think it’s good marketing, but it’s not true when it comes to McNealy’s statements about cloneability. It isn’t.
You cannot take an application developed for a Sun and easily port it to anything else. I don’t care if it uses the Satie UNIX and NeWS and the whole bit, because there are other ways it gets subtly wound around Sun hardware. So they have a captive audience. They absolutely have people right by the nape of the neck. It’s not as bad as someone being wired into VMS on a DEC machine where they’ve written some assembly code, but it’s still an insidious kind of commitment that almost catches people unawares. So it’s not true, it’s not real, this notion of cloneability and they’re totally open and all of this—they’re smart enough to realize that people are wired into them, and they put out this image, and it’s been extremely successful. But it isn’t true.
Well, from the user standpoint the cloneability of the PC was a great thing, but it sure horrified IBM once they realized what they’d done.
And in fact they have given up on that now, haven’t they? And they’re getting criticized for going to something that’s somewhat more closed, but basically if Sun really believes in cloneability, then make their stuff available at low cost. License it to Sony, for example, or NEC—that would scare them, and for good reason. So I think that Sun’s got a great marketing campaign going, but in reality anybody who’s ever written a piece of code for a machine realizes that—you go to any software developer, and if they have to change one line of code in order to make it run on a different machine, it’s a massive decision. Because for one thing, there may be a million lines of code and they don’t know where that one line of code is, you know? And they’ve got to go in and find it and change it, maybe. It’s a big decision.
But isn’t the open architecture to some extent an outgrowth of the portability of UNIX?
No. UNIX was set out with the true objective of being a portable operating system. But not for any marketing reasons, just for technical reasons, because the people at Bell Labs who developed it wanted to be able to run it on different machines. It was intended to be portable. Mainly it ran on DEC machines.
See, the fallacy is that Sun’s marketing of the openness is believed by people, and Sun maybe even believes it themselves. But marketing is a large part of the game, and they’ve been a very successful marketer.
Our company is—there’s some differences. They appealed a lot to computer science types. We appeal a lot to non-computer science types—the artists. We appeal a lot to the artists. A lot to the physicists, the chemists, the medical doctors. They see our stuff and say “That’s something I can relate to! Gee, is there some way I can do this?” And the animation people, and mechanical engineers. The analyst—the person who’s trying to understand why a part might fail under a certain scenario—he’s trying to see the part. Finite element analysis. Fluid dynamics. Studying an airplane flowing through the air. These are all three-dimensional pheonomena that people are trying to understand and deal with. Those are the kinds of people we appeal to, much more so than to the computer science, the program developer for its own sake.
Would that be at all analogous to what the Mac did in the personal computer world?
Yes. I think we’re very similar. I think there’s a very analogous line to draw between ourselves and Apple. We’re very similar in that we offer these visual programming environments. It’s a graphic computer. And we will expand into other more conventional 2D and image processing areas, because we want to be a completely integrated graphic computing environment above the low-end distribution channels. Above the sort of $10,000 price range.
The DOS computing world doesn’t know how to take Sun at all. It’s pretty funny to see an audience of Suits thrown into complete confusion by a typical Sun “here it is, here’s why we’re doing it, take it or leave it, bye” presentation.
Well, on a technical basis the Sun machine is better than a PC by any measure. Their low end machine is a better machine than a PC or a Mac, on a technical basis. But on a technical basis, a Silicon Graphics is better than a Sun machine. It is. It’s got better graphics, a faster CPU.
But there’s a very different set of marketing things, and Sun of all people ought to know the value of being there first at the low end. See, Apple and IBM and Compaq own those low-end distribution channels, and Sun is having a real hard time breaking into them. Even if they do, I think the low end is a very, very simple community compared to the community that Sun has been selling to, simple in the way they think about computing. UNIX is overwhelming to them. I’m talking about the people that work at Businessland and sell computers—go in and talk to those people and see if you think they’re ever going to be able to understand UNIX. No way.
I’ve talked to salespeople who have trouble understanding MacWrite.
(Laughs) That’s right.
I mean, they’re ready to take your check and pull out the order and check off the features that you want, and some of them have some reasonable ability to articulate what’s in it—they never even use the program.
It’s a very different problem that Sun is trying to address as they go down there. But on the other hand, I do believe that they have a better product.
With the Personal Iris, you’re going into new distribution channels yourselves, aren’t you?
Well, not really. To some degree that’s true in some of the dealer development. We’re getting into some of the channels that companies such as Sun have had going for a while, and we’re being reasonably successful at that, but by and large it’s still using our direct sales force and selling larger numbers to our big current accounts and to new accounts as they open up.
We are now at a point where Sun was in price a couple of years ago. I’d like to say something about price, though—a lot of people say, because we're at the same price, “oh, you must be competing with Sun,” or “you must be competing with so-and-so” All I have to do is point out that you can buy a $15,000 motorcycle or a $15,000 pickup truck—they’re not the same product, they’re not the same market. We have a $15,000 3D machine, and it’s going to have some overlapping capabilities with a 2D machine, such as the one Sun sells. But they’re really very different products.
Eventually, I think there’l] be some conflict in the market, but at the present time we don’t see much.
How are you planning to keep your edge?
Aggressively developing software packages that really use 3D so that we offer something that’s distinctly visibly different. That’s the nice thing about graphics, it’s visible. Distinctive. So the key is getting people to make the transition from the flat earth society into the world of 3D.
Is there any overlap between the NeXT’s market and yours?
Personally, I think the NeXT has just sort of come along too late. It’ll be a modestly successful company, but I don’t think there’s any way it would ever be a successful enough company to be a real threat to any of the current workstation companies or personal computer companies.
Do you agree with Scott McNealy that it was a bad idea not to make it a RISC machine?
Yeah. I agree with everything he said there, actually, in that particular line. Just a bunch of mistakes. I mean, why use a non-standard operating system?
And the 68030 is relatively old news at this point.
Oh yeah, that was a bad mistake. And part of the reason for Mach was that [Jobs] viewed it as something that fit what he saw as a need—it kind of fits in the whole scheme of the object-oriented world. And it’s a university-based system, and ostensibly he’s after the university market, so he says. Of course he’s also smart enough to realize that there’s a lot of follow-on from the university market.
Sure. Look at the Macintosh.
Yes. But there was just mistake after mistake. Probably the biggest one being, being so late.
And it’s not done being late yet.
I think he understimated (sic) the amount of work required to get it there, the amount of money that was going to be required to get it there, and there’s no reason why, if Sun does the job that they should do, they shouldn’t he totally blow him out of the water.
I don’t see him as a threat to us. He’s got some designs on some aspects of what we do. He sees the value of what we do—Steve’s a pretty insightful guy, he sees what’s going on here. But his ability to pull off that plus all of the other things he’s doing is—I’d say it’s very unlikely.
We don’t think about NeXT. We don’t worry about NeXT. Hardly ever think about them any more, especially since he isn’t shipping. Once he starts shipping, I think there'll be something you can look at.
Steve’s name is only going to carry him so far. The rubber hits the road someplace out there. (laughs)
What computer ads would you like to see in the future? Please comment below. If you enjoyed it, please share it with your friends and relatives. Thank you.