Workstation Leadership Forum 99

Craig Barrett
Burlingame, Calif., USA
June 30, 1999

I guess in a couple of years, we'll have to move it to Las Vegas and replace Comdex.

The other introductory comment I want to make is to thank the Congress and the President for deciding to pass a Y2000 bill so some of us who have been spending most of our time lobbying with them in that direction can now get back to doing some more useful work.

I do want to talk this morning about vision and the big picture of what's going on in this space, not only in just workstations, but enterprise computing in general, and put that in really the perspective of a worldwide vision, what's happening with the Internet and connectivity amongst computers.

And many of us in the computer industry have a very simple vision, that we're looking at something of the order of a billion connected computers in the next five to six years, connected around the world. And this means a connectivity within corporations, which has been important in the past, gets to be more important as you now want to have everyone in your firm connected not only to everyone else in the firm, but connected to the outside world, customers, vendors, suppliers, et cetera.

And the strategic inflection point we're facing with this exponential growth of the Internet is having an impact on all of us as we talk about computer infrastructures, applications of the future, cost of ownership of our computing infrastructure.

So from a vision standpoint, I want to really outline a few things to talk about today. One is, the convergence of workstations into the enterprise system is being accelerated by this basic building block that we're here talking about today, that is, Intel Architecture/Windows NT-based computers. We're able to integrate these systems in with the rest of our computer infrastructure and bring our knowledge workers really into the mainstream of the computer backbone.

Second, that Internets, intranets, this concept of connected computers around the world, where it's necessary to share data to communicate instantaneously, to communicate with all sorts of people around the world, next door, next continent, is important. So the Internet utilization is fueling this need for convergence.

Third is, from a numerical standpoint, the IA/NT-based workstations are really the platforms of choice. And we'll show a bit of data a little bit later on in terms of the numerical superiority, if you will, of this platform as it continues to grow in importance relative to the competition.

And, lastly, I'll make a few comments about the 64-bit architecture which is going to be with us next year and where we think that that architecture will play a role in workstation computing and some of the excitement that we have about that pending introduction.

Now, if we look very simply at this concept of convergence, and, really, it's the convergence of bringing knowledge workers, those people who have been doing design, analysis, whether it's financial analysis or engineering analysis or content creation, into the mainstream, it's really making sure that the workstations are connected to the corporate backbone.

And if you look at, really, what we call enterprise without convergence, we see what, historically, has happened. Workstations, with their compute servers, were on one system, and everyone else in the corporation, business desktops, were really on the enterprise server on another system.

It's kind of interesting, because quite often, the first network connectivity within our organizations really came out of the engineering environment. But the engineering environment is quite often one of the last to be physically connected into the enterprise backbone.

If you look at what we're talking about with convergence, it's really connecting the workstations, compute servers, with the enterprise servers so we have a seamless system. We have one desktop machine for our engineering or knowledge workers that can adequately communicate with the rest of the corporation, we've got data-sharing, we have a lower cost of ownership because of common architecture.

And if you throw the Internet into this system, with Internet servers inside and outside, this really forces the convergence from the standpoint of the necessity to communicate with people on the outside, share data, share services.

The whole issue of the Internet is Internet time or time to money, speed with which you can get your product and communication done, and at a lower cost. So this whole issue of competitiveness really drives the convergence of the engineering backbone into the corporate backbone and into the Internet backbone.

Now, when I was in front of the group last year, we talked a little bit about what Intel was doing along these directions, just from the standpoint that we wanted to demonstrate that we were following our own advice. And I wanted to give you a quick status report on where we are relative to introducing this workstation backbone and integrating it with the rest of our system.

If you look at what we've done in desktops, mobile, and servers, in fact, this is completely done in terms of having all of these on IA/Windows backbone. But I think, more interestingly, it's what we're doing relative to bringing the knowledge workers in our company along on this backbone.

And workstations which are involved in board design or electromechanical system level design, this is completely done. This is an IA/NT backbone. One workstation per desktop for those knowledge workers, fully integrated with the rest of the company.

If you look at what we're doing from an integrated circuit design capability, we're making great progress here. Much of our capability has been moved over to IA/NT. And I don't want to talk in great detail about this, because when Bill follows me, his keynote presentation, Greg Spirakis, who's the head of our design technology at Intel, is going to be one of his demonstrations. And I'll really let Greg come up and talk in detail about what we are doing here. But we are making strong progress in this space.

And Intel and Microsoft, I think, are working very well to improve the interoperability as we move from Unix-based capability over into IA/NT-based. And you'll hear some of this in the workshop tomorrow and hear other people present on this topic.

So I think, overall, we're making very good progress in this convergence within our own company.

Now, rather than just talk about Intel, I also want to talk about other people and the convergence that's occurring there.

I thought the first example we might pick out would be one of mechanical design and CAD capability. And what I want to do is show you a brief video from Daemler-Chrysler {?} and show you what they're doing in this space. And then we'll follow that up with a little bit more detail and demonstration.

So please roll the video.

(Video playing.)

VIDEO: We're changing the way cars are created. The new Dodge Intrepid is the world's first car designed, assembled, and proven in a digital world, absolute mathematical precision, a world where billions of ones and zeroes add up to higher levels of refinement, where new ideas grow on logic trees and where digital roads lead to virtual perfection.

Welcome to the world of the new Dodge Intrepid. We're changing everything again.

CRAIG BARRETT: I like that last tag line, "changing everything again."

There's an immense amount of processing power that goes into that system to create those models that we talked about. And besides the processing power, there's some great software behind it. That's why I've invited Bernard Charles, who's president of Dassault Systems, who creates much of the software that goes into those models and images we've seen.

Bernard, why don't you tell us a little about Dassault while you're up here.

BERNARD CHARLES: Nice meeting you, Craig. Thank you for inviting me here.

Good morning. Bonjour.

Dassault Systems is in the mechanical design information market. We have five (inaudible). The most precise one is Catia {?}.

The video you have just seen here has been generated with Catia from Catia data. We have also, (inaudible) as enterprise information system on the solid works is known for the entry solid modeling on PCs, I would say. And we have, finally, (inaudible) for a system. It's 2500 people, an alliance with IBM to sell the solution worldwide.

And we are basically one billion dollar end user revenue. We grew about 30 percent last year.

CRAIG BARRETT: I don't want this to be totally a Dessault commercial. You can tell us how IA/NT workstations have played into this and how they're --

BERNARD CHARLES: Absolutely.

By the way, if you want to order Catia, we have some outside.

The theme of convergence is a very critical element. Our vision is the digital enterprise. And your presentation, clearly, we applaud this initiative of substantial integration across the company.

The digital enterprise, in short, it's the (inaudible) for customers to do integration of the whole product life cycle, from the initial idea to the product in operation.

You can see that for cars. You see that for automotive, for the entire automotive segment, but also for (inaudible) space. I used to say that, "The world flies with Catia." Most of the airplanes in the world are done with Catia today. On the new platform, IA/NT platform, the goal is to make this viable for many other segments, not only small-, medium-sized companies because of the scalability, but also to large corporations for the integration.

CRAIG BARRETT: Great.

BERNARD CHARLES: So, for that, we have created a new architecture to take advantage of the Windows environment.

And, basically, this is a good illustration of how important this convergence is going to be for the customers.

This Version 5 architecture contains five things. First of all, these Windows Web metaphor, which (inaudible) integration.

The second thing is the scalability, which is very, very critical.

The third element is related to the knowledge. Because, of course, all of those products, the planes, cars, you have a lot of information which is related to the product knowledge which need to be made available to everyone across the company, inside the company, as well as suppliers.

And, clearly, the value of IA/NT is to make this simple, simple for the end user, simple for the deployment, and still get the (inaudible).

CRAIG BARRETT: Great.

Maybe what we can do is, rather than just talk about V5 and Version 5, maybe you can give us some example of how that's used in places like Daemler-Chrysler.

BERNARD CHARLES: Absolutely.

There is a good illustration where Daemler-Chryler is looking at deploying the digitalized model for all functions, not just engineering and manufacturing, procurement, et cetera, all those kinds of functions.

And here is, I think, a little video that shows how they plan to do it.

VIDEO: (Inaudible.)

Here at Chrysler we are working very diligently to streamline our (inaudible) process. We've organized new platform teams. We've removed a lot of the bureaucracy. We've established new relationships with our standard enterprise. With Dassault moving the continuing family of products to the Intel and Microsoft/Windows platform, we now can extend our reach to a number of people throughout the extended product (inaudible) enterprise that heretofore were left out.

We have IBM RS6000s {?} and SGIs. We have to provide system administration and Catia administration on both of those platforms. With the NT platform coming up, we're hoping in the long term we get to a point where we have only one platform that we have to manage.

The problem we've had in the past is that there's people in tiers two, three, et cetera, that just didn't have access to this kind of capability at an affordable price.

We're very pleased with the way Catia runs on the Intel Architecture/Windows NT platform. We're excited about the new version and the functionality that it provides us, (inaudible) really slick, speedy, and yet passionate way of making these cars and trucks that people want to buy and enjoy driving and want to buy again.

CRAIG BARRETT: I like these videos, but I think what you ought to do is show us a real-live demo that's a little bit of risk-taking.

BERNARD CHARLES: You want sometime to use the product, but not today, I understand.

Craig, we have a demonstration here, basically, which shows on the screen here a Concord, a Chrysler Concord car.

This is running on an NT (inaudible), a dual processor (inaudible) 553 {?} MHz. And as you can see, we have, basically, for this kind of digital mockup, you have about 15 gigabyte of data. As a side note, I will say the triple seven {?} is three terabytes. So a lot of data.

Here in this navigation, you have approximately 1.5 gigabyte. And it's really not only the shapes, but you do have on this platform the entire product behavior. So this means that you can really look at every part, you can simulate the behavior. You can even, at any point in time, open the door or do some basic changes.

CRAIG BARRETT: Great. Can we open the door?

BERNARD CHARLES: Sure. And you might even try to drive the car maybe one day. Not today.

CRAIG BARRETT: Not today.

BERNARD CHARLES: So we can just -- here we have what you call edit in place, this Windows environment that allows you to really change the abstraction and go extremely quickly and play with the door realtime.

As you can see, it's extremely easy to use. A few clicks.

And you can even do some analysis on the fly. So if you want to do analysis on the fly, you click on the part. And here is something that was just done before this demo, a simulation of the hinge being presented on the Internet.

In short, Craig, what we want to do with this platform is to do what we call CNN live, to provide to everyone across, inside the company and outside the Catia News Network.

CRAIG BARRETT: Great. I think that really demonstrates the ability to communicate inside and outside the company, across the board, with your suppliers, with outside engineers, as well as your own inside engineers.

Great demo. Thanks, Bernard.

BERNARD CHARLES: Thank you very much.

(Applause.)

CRAIG BARRETT: I think it's applications like that and their rapid growth which is really fueling this sort of growth we're seeing in the overall workstation segment.

And what I've tried to show on the slide that's on the screen now is, basically, a forecast, and this is from IDC data, for the last couple of years and for the next several years in terms of workstation growth.

My personal feeling is that these growths are underestimated on the basis of how important content creation is going to be for the Internet and its explosive growth. But even if you look at the numbers that are here, which are 20 percent compound growth over the next several years in terms of the overall growth, but, more importantly, Intel Architecture and predominantly IA with NT, is in fact growing substantially in this time frame. And the older, established RISC architectures are, in fact, having a negative growth rate.

I think it's also interesting to note that if you look at what was driving the growth over the last couple of years was basically the concept of a single platform or a single footprint on the engineering desktop so that you could tie them into the corporate network with a single instrument, and also the price-performance.

But if we look forward, it's actually performance plus price-performance plus cost of ownership which is driving this capability.

Now, just looking at the improvements from the Intel perspective, since I talked to this group last year, I think if you look at all possible measures of processor capability, processor performance, whether it be the generation of processor, the frequency that the processor's running, the memory addressability, the graphics capability, or the system bus performance, we're showing improvement across the board. And the upward direction of this arrow is indicative that we expect to continue to see this fantastic improvement as we go forward.

Now, if you look at this just from a performance standpoint, whereas we were selling price-performance before, today we're selling a combination of performance, price-performance, and cost of ownership.

I tried to show here some Pro/ENGINEER benchmarks in terms of the Intel Architecture-based systems from a number of OEMs, which is the left-hand curve. And we're really, in this space, talking about leapfrogging our constant battle for the top performance capability between RISC-based systems and Intel/NT-based systems.

But if you look at the price-performance, and whereas we see that, from a performance standpoint, we're in a relatively equal basis, price-performance, the IA-based systems are in fact three to five times or so more price-performance competitive, price per unit of workstation horsepower, than the other systems.

The fact that you also can get these systems from a variety of OEMs gives the end user more choice in terms of capability, more choice in terms of supplier.

Now, if we very quickly look at what's happening in the support capability here, as we go through workstation growth over the next several years, and not just from Intel's perspective here, and Intel is obviously doing a number of things in terms of the processor development, workstation, or the chipsets that go with the processor, which are increasingly important in overall workstation performance, but boards and chassis from our workstations products group that Anand, who was up here earlier, runs. And also specs and guidelines, whether it be AGP Pro, the specs for the next-generation graphics capability, whether it be the specs in PC 2001 for workstation specs in that time frame, we're involved in the basic architecture and the innovation, with many others, including Microsoft, to provide all of this capability into the marketplace.

So from an architectural innovation standpoint, we continue to move forward.

If you look at what's happened from an OEM standpoint, there's been an absolute explosion in number of people providing these high-performance, low-cost workstations.

And initial introduction back in the '96 time frame, Intergraph was the first really to make this conversion, but shortly thereafter, followed by, essentially, all of the major computer OEMs around the world. And then in the last 18 months or so, this has been a real avalanche of people involved here.

You can see, essentially, all of these suppliers at the Tech Showcase. And I encourage you to stop by their booths and chat with them about their offerings.

From a software standpoint, I mean, there's an equal explosion in terms of whether you're talking about mechanical design, content creation, electrical design, finance, or looking at the major challenges in the geological or scientific arena, we have a wide variety of applications here. And this continues to grow as more and more people port their applications onto this IA/NT backbone. Again, all of these are going to be, or many of these will be present at the Tech Showcase as well.

Now, there's not only people just porting their applications to this architecture in private, but, in fact, in conjunction with many of the OEMs, we have set up application solution centers. And I've really tried to show a collection of them here around the world. And these are basically solution centers designed to provide porting and software tuning capability on IA/NT architecture backbone.

Now, the black boxes here really represent those which are targeted more towards workstations, and the red or pink ones are more toward the server space as we go forward.

The point is that we're getting more and more capability to tune applications to this architecture to speed up the acceptance of the architecture as we go forward.

And, actually, I think Richard Wert {?} from Intel will be on stage tomorrow to discuss this in some detail as well. So I will not try to steal too much of his thunder.

The important point to note is, these are done in conjunction with major OEMs who set up these centers and then to fine-detail the hardware/software capability as you tune and port the applications.

Now, graphics technology, I think, is just as impressive in terms of all the major graphics ISVs are (inaudible) supporting the architecture. We've shown the majority of them on this slide. Not only are they supporting the IA-32 architecture, but also many of those on this slide are supporting the IA-64 architecture with public commitments and starting to generate the hardware and capability as the Merced architecture comes out next year.

Along the lines of the graphics capability, I do want to make one announcement which I think came out in a press release this morning in terms of Intel and SGI focusing on an optimization agreement to take Open GL and Version 1.2 and target it for the Intel Architecture for the very high-performance workstations.

I think it's announcements of this sort which will help, in fact, lift up the entire capability of the workstation/visual graphics industry from a performance standpoint and make it even more competitive in the marketplace.

This initial optimization will be focused, obviously, on the Pentium III, Pentium III/Xeon families. But we're looking forward as we go to the next generation, IA-64 families, working with companies like SGI as well for optimization in that space.

Now, we could talk about visual graphics and content creation, and I could show you slides. But I think perhaps one of the best ways to demonstrate this is not through my talking about it, but to have another physical demonstration here on the stage. And this time, we would really like to talk about digital content creation.

And digital content creation, I think, is important for two reasons. One is, the Internet grows -- as the Internet grows and Internet commerce grows, it's going to be driven by digital content creation, which will be the backbone of our visual experience with the Internet.

But if you look at the entertainment industry, it's almost equally exciting there. As we go to movies today, we can't possibly tell what was real and what was created off of high-performance workstations. And so the content that we see on a daily basis in an entertainment mode is increasingly digitally created in character.

And along these lines, what I'd like to do is introduce Frank Foster, who is in fact the founder and senior vice president at Sony Picture Imageworks. And Frank is going to come out from somewhere -- there you are. Hi, Frank -- and talk to us a little bit about what you can do from a digital content creation standpoint.

FRANK FOSTER: Well, I'd like to explain a little bit about what's going on in Hollywood with the migration. I think most of the facilities in Hollywood have either moved to Intel/NT or are trying to figure out how to move.

CRAIG BARRETT: I love that.

FRANK FOSTER: The bigger the company, the older the company, the slower the migration. But it's all under way.

To be more specific about what we're doing at Sony, I'd like to show you one of the things we're working on currently and tell you a little bit more about the migration that we're going through.

VIDEO: The little family is about to do -- "I want a little brother, not a big brother" -- something perfectly extraordinary.

"Mr. and Mrs. Little, we try to discourage couples from adopting children outside their own species."

"Hello, everyone."

This holiday season, Columbia Pictures and the director of "The Lion King" invite you to meet Stuart. Stuart. (inaudible).

"Do not eat family members."

Stuart Little.

"Bad cat."

"Come out with it."

This film is not yet rated.

CRAIG BARRETT: I think it's only appropriate that you have a mouse for a character before this group. But that's okay.

FRANK FOSTER: Well, you can imagine when you've got a movie where the main character is a CG, generated character, that it's going to be an expensive process. So we've been looking for ways to help us cut costs. And, in actuality, we've been using Intel- and Microsoft-based products for quite some time.

But on this particular movie, which we're still in production on, we have switched our complete digital daily system over to Intel/NT. And when I say "digital dailies," in the old days of Hollywood, you'd shoot your film, go to the lab. The next morning, the director would view it. And we called it dailies.

In our process, it's similar except we don't go to the lab. The animators are all working on their shots. They render them over the night, and they get compressed and shown on a system.

We used to do it manually. Today, we've worked with several of the third-party companies. And I'm showing you actually the system we use for digital dailies right here. Instead of using videotape and the older process of going directly to a small, hard dedicated system, we're using the VPS perception card with the Insynch Speed Racer {?} software, which is actually an editing program.

I told you the process can be a long migration process for some companies. We actually started doing this in 1992, where we brought in 486s to help us do electronic storyboarding.

This was for our Bruce Willis film where we are using the computer to help us create electronic storyboards that are architecturally accurate. In this case, we took the plans from the Pittsburgh city planning department and actually did the chase sequence completely on computer shot by shot in advance so the director saw the movie before we ever shot a frame of film.

That was fairly new back in 1992. But today, almost every major film, especially the ones with expensive shots, are using this kind of process.

You can see how the real footage and the previsualization footage actually match up, even on this project we did quite a while ago.

CRAIG BARRETT: I can almost remember 486s. (Laughter.)

FRANK FOSTER: And I can almost remember DOS. (Laughter.)

FRANK FOSTER: In 1995, we took it a step further. Again, this was with DOS-based systems. This was 3D Studio we're using. This is Johnny Mnemonic {?}.

We worked with William Gibson, who coined the term "cyberspace" to create this sequence, where we are looking at the Internet as it may look 40 years in the future.

CRAIG BARRETT: We get the point. That was some good simulation there. I don't think the movie was quite as good as the simulation, though, was it?

FRANK FOSTER: Craig, you're very right about that. (Laughter.)

FRANK FOSTER: But, fortunately, we made an impression on the producers. They looked at the cost, they looked at the quality, and we got a chance to work on another film, this one for Columbia, a few years later. But this time, we're using Intel and Microsoft to do full-screen photorealistic character animation, in this case, hundreds of butterflies.

I'll show you how we're doing it.

We have wire frames that are kind of rotoscoped or matched to the actresses in the scene. We created separate layers for depth of field or the out-of-focus butterflies that are out of the shot, doing a Gaussian blur on them.

Using the digital butterflies in the light, we projected the shadows down onto those meshes so we could get accurate distortion of the shadows that we then subtracted in the background plate, which was the photo plate. And we produced these two shots.

We did two shots in the film on Intel/ NT-based system. And, ironically, those are the two shots that the director loved the most in the movie and were the cheapest to do.

CRAIG BARRETT: Great. Well, what else can you show us here?

FRANK FOSTER: Well, I showed you a little bit of "Stuart Little," which is our upcoming holiday release. We're still in production on that.

Let's show a little bit behind the scenes of how we're doing this.

We've migrated about 50 percent of our digital character animation over to Intel so far. Here you can see an animator working with the mouse mesh. Had to create our own fur.

We actually had our software developers having meetings with costume designers on how to properly costume our characters, which, of course, had to be generated digitally.

We have a whole department devoted specifically to character animation. And we almost refer to these animators as actors. They have to put the performance into the character.

And, finally, if you want, I'll show you a section from the film that no one's seen yet, but I have permission to show in this group.

CRAIG BARRETT: Great. Well, as long as everyone signs a nondisclosure out there, we can take a look at this.

FRANK FOSTER: This is a little sequence we call the washing machine, which will give you kind of an idea of how the digital material intercuts with the live action.

VIDEO: Okay.

"Thank you."

"Oh, dear." (Laughter.)

VIDEO: "Mom! Mom! Hello, Mom! It's Stuart. I'm in the washing machine. Mom!"

"Where are you going?"

"Hi, Stuart."

"Mom!"

"Stuart!"

(Coughing.)

"Stuart, are you all right?"

"I'm okay, Mom." (Laughter.) (Applause.)

CRAIG BARRETT: Thanks, Frank.

FRANK FOSTER: Thank you, Craig. (Applause.)

CRAIG BARRETT: I think the personality of some of their animators matches those of some of our architecture designers as well. (Laughter.)

CRAIG BARRETT: If you look at what's coming in the future in terms of performance, and I think performance is important to most of you out there because it just is going to magnify the capability with the architecture that we're talking about today, we're really in the 500- to 600-MHz range with our 32-bit processors today. The industry and Intel are moving to the next generation of process technology, .18 micron. And then there are four or five other generations moving down after that that we already have on the drawing boards, should be in the gigahertz or thousand megahertz or .001 terahertz, depending on your inclination here, by the end of next year in terms of our processor performance in the 32-bit family.

So Moore's Law, which we've been faithfully following for the last 35 years, looks like it's good for another 15 or so in terms of doubling performance for you every 18 months.

And not only from the standpoint of the 32-bit family, but we're also introducing the 64-bit family the second half of next year, the Merced family. And we expect to see an added boost there in terms of 64-bit compute capability and some added features for their very high-performance workstation and server applications in terms of scalability and reliability and absolute compute crunch power performance.

Very excited about what's coming there. Very excited about the whole IA-64 family.

And, in fact, if you look at what we're doing in that space, we think that from a 64-bit computing platform, this architecture stands the high probability of just consolidating the industry and getting people moving in this space for new and exciting performance.

A number of companies are already committed into this space to go forward with us in the year 2000 introduction. And one of the things we've done to try to facilitate this is to create an investment fund of the so-called Intel-64 Fund, which is a quarter-billion-dollar investment fund. It's actually small by some of Microsoft's investment standards, but reasonably sized from an Intel investment standpoint.

But Morgan Stanley is the lead investment entity here, with a number, essentially, most of the major computer OEMs, and then companies like Enron, Bank of America, Ford, et cetera, users of the technology. So combination of OEMs, end users, Intel, involved to, in fact, create a fund to fund both server and workstation applications on the IA-64 family to bring those into the marketplace more rapidly.

Very excited about the possibilities which are going to happen there.

And, in fact, just looking at the Merced processor itself, we're very excited about it as the first instantiation of this new architecture. I don't want to dwell too much on Merced, because it is just the first generation.

There will be code name McKinley follow-on to that. And then there are several other generations which are already under architectural design beyond McKinley.

Merced will be out in mid-2000. We'll be sampling the product within the next two months or so, fully binary compatible with all the IA-32 capability that's in the marketplace today.

A number of operating systems, including Win 64, are already booting on the Merced simulator. And while, again, I could talk about this, it's probably more interesting to show you what you can do in this space in terms of simulation of what's going to be on this architecture.

We do have the simulator for it. We do have software development kits from Microsoft. We do have people who are in fact porting their software onto the development system as it stands today.

What I'd like to do here is in fact do a very simple demo to show you what can be done and to show you how ready the system is to accept these new applications.

So Kenny Quan {?} from Intel is going to come out. And we're basically going to do a very simple demo of what's possible on a simulated IA-64.

And we're going to talk about Mental Ray {?}, I think? Is the application, Kenny; right?

KENNY QUAN: That's correct, Craig.

Okay. Here we go. And let me briefly describe Mental Ray. Mental Ray is a photorealistic rendering software that is used in many high-end digital content creation applications such as SoftImagine V3, Sumatra {?}, Catia, and 3D Studio Max. They currently consist of about 500,000 lines of C code, which took our developers three days to compile and get it up and running on our simulator. And it uses an incredibly photorealistic image, which is quite good quality as well as geometrically accurate as well.

CRAIG BARRETT: So you're not going to show us any realtime simulations today?

KENNY QUAN: Unfortunately, not today, Craig.

CRAIG BARRETT: This thing runs, what, probably at the speed of a 286?

KENNY QUAN: Yes. And that wouldn't be too exciting for the audience out there today.

CRAIG BARRETT: This is like nostalgia. We've heard about 486s and now 286s.

But I think the point here that you're trying to emphasize is that the software application runs on the hardware simulator. It's ready to go as soon as the actual hardware is available.

KENNY QUAN: That's exactly true. And, in fact, Mental Ray is going to be one of the first applications to actually run on the Merced processor when that becomes available.

CRAIG BARRETT: Great. Thanks, Kenny.

KENNY QUAN: Thank you.

CRAIG BARRETT: We think the IA-64 family is going to coexist with the 32-bit family for an extended period of time. And some of the exciting initial applications in the 64-bit space will be in fact those areas which get maximum benefit or pay-back from the IA-64 characteristics. And some of these will be finite analysis and high-end digital content creation and financial risk analysis and synthesis and design rule-checking.

There will be other areas which are lower-performance, content creation and mechanical desktop design capability and schematic capture. Some of those things will come later.

We're very excited about the large number of applications which we think will be present with the architecture when it hits the market.

And if you look at just some of the many, many people who are already committed to this architecture, both from a OEM standpoint, from an ISV standpoint, from a software standpoint, I think it's a very, very impressive list of who's who in the industry.

So we're very excited about the addition of this generation.

And next year, when we have the same conference, and I hope there are four times as many people in it as we have this year, we'll have IA-64 up here on the stage and be doing demonstrations with it and showing the great capability that it will bring to the marketplace.

Let me conclude by emphasizing that this IA/NT workstation capability I think is one of the great unifying influences. It's allowing us to converge our knowledge workers into the rest of the compute backbone not only inside of our companies, but outside of our companies, as the Internet continues to be a great driving force for this.

And if you look at the volume workstation driver today, with about five times the sales rate in terms of units into the workstation marketplace, it's clearly IA/NT-based systems. The IDC forecast I showed you earlier showed numerical superiority. And in terms of a dollar-weighted sales standpoint, they're forecasting sometime this year or early next year that the IA-based workstations will, in fact, from a dollar standpoint exceed those from the other architectures.

I think this carries two bits of information with it, the increasing importance of IA/NT in that space, but also the great price-performance it brings from the standpoint of putting five times the number of workstations at about the same total cost into the marketplace, thereby increasing your cost effectiveness.

It is a no-compromise platform solution. That is, it's got a phenomenal number of OEMs supporting it, a phenomenal amount of software supporting it. And it in fact gives full capability to those knowledge workers to bring them into the environment.

And, lastly, we are excited about the IA-64 family. And it's going to bring the continuing benefits and economics that we've seen from IA at the 32-bit level, and looking back even farther, at the 16-bit level, into this space.

So we think, from a workstation server standpoint, we're looking at a whole new era of price-performance capability as we go into the IA-64 area.

I think perhaps I've talked long enough in this space, and what I really ought to do is turn the podium over to my counterpart from Microsoft, Bill Gates. (Applause.)

(Gates keynote.)

Q&A WITH CRAIG BARRETT AND BILL GATES

ANAND CHANDRASEKHER: We have two microphones set up, one along this side over here, right there, and one along this aisle over here. If you have questions, if could you line up behind those mikes, I'll alternate between the two sides.

One comment. The topic for today is Workstation Leadership Forum. So -- (Laughter.)

ANAND CHANDRASEKHER: So if you could focus your questions on that specific subject matter, that would be great.

Q. From Audience: My name (inaudible). I would like to know, it seems the computer is getting faster and faster. However, last year, you mentioned you would like to do something for the computer against (inaudible) provide the technology that's (inaudible). How do we work with you?

BILL GATES: You're a software developer?

Q. From Audience: Yes, we are.

BILL GATES: Well, we're definitely interested in talking with you about what you're developing. And I'll have our developer relationships group get in touch with you.

Q. From Audience: Is there any way I can meet you? (Laughter.)

BILL GATES: I've got a great staff. And let's figure out how we can work together.

ANAND CHANDRASEKHER: Question from this side.

Q. From Audience: I was going to ask Craig about the floating-point calculation unit in the Merced processors or from Pentium III to Merced. Are you going to improve or add more floating-point capability?

Thank you.

CRAIG BARRETT: Yeah. The Merced has really great improvement over the 32-bit family going forward. And McKinley, beyond that, will be even a new microarchitectural change, probably 2X Merced performance. So we're very excited about floating-point capability of the whole 64-bit family going forward.

ANAND CHANDRASEKHER: Question here.

Q. From Audience: Yes. With more and more people due to high-speed Internet access from the home working from home, with the newer workstations that are going to be coming out, will they also be usable by the home user for their home applications for their children and such?

I've noticed that if we have a workstation at home, a lot of times, it's higher-performance, especially on the NT-based, that they're having -- a lot of the programming is not designed or will not operate. They're still able to work from home, but they're winding up with two or more PCs.

BILL GATES: Well, that's a very good question, because the whole goal of having a common PC architecture and Windows in common between them is that it's really seamless. And, in fact, when we're doing Windows 2000 reviews, we go from talking about how many games run on Windows 2000 to does Catia run well on Windows 2000, because we want to address that full range.

If you go back in time, a lot of the entertainment kids' software was specifically only designed for and tested on Windows 95. We feel like we've made an incredible amount of progress, and there's a lot of things in Windows 2000 to make sure that that technology, the Windows NT kernel not only runs those applications, but that we actually can use that technology even in consumer versions of Windows in the years ahead.

So we're moving off of the Win 9X code base to that NT code base.

And so if you have any applications that aren't working on NT, let us know, because we think we've covered that very, very well with Windows 2000, making it have the consumer compatibility and the high-end applications compatibility.

So if you want to have two PCs, that's fine. But all the software should run on both of them.

ANAND CHANDRASEKHER: Question from this side.

Q. From Audience: Yeah. Last year's forum, you discussed quite a bit about the Fahrenheit initiative. And I haven't heard anything about that today.

Craig's slide also had a new initiative that you've got with Silicon Graphics to do Open GL. Could you guys comment on the status and how this relationship would affect that Fahrenheit.

ANAND CHANDRASEKHER: Craig, do you want to go first?

Do you want to go first on this one?

CRAIG BARRETT: No. (Laughter.)

BILL GATES: There is a, really, three-way collaboration, where Intel and Microsoft and SGI are both taking Open GL as it exists today and doing better and better implementations of that.

The new version of that, you can say, is under this Fahrenheit initiative. And, actually, the graphics primitives that Fahrenheit addresses actually go even higher level than Open GL has. There's continued good work on Fahrenheit.

And I'd say that we all have a goal that at the Workstation Leadership Forum a year from now, you should see those efforts.

CRAIG BARRETT: I concur. I think we're all working together just to raise the bar, raise the level of performance across the board.

ANAND CHANDRASEKHER: Question from this side of the room.

Q. From Audience: Yes. Thank you. This is for Mr. Gates.

Many of us have legacy systems running on Unix with relatively large customer bases. One of the things that my company's been involved in is a, actually, very massive effort based on COM {?}. It's a technology that we absolutely are in love with.

But one of our issues is about how we move legacy users from Unix in a system that really doesn't provide much in terms of development suite like COM does.

I'm kind of curious about your directions for COM interoperability in the Unix environment, if you could comment on that.

BILL GATES: Sure. Well, COM itself is sort of a way of components talking to each other. And so it allows components to describe what capabilities they have. And you get the kind of replaceability that object-oriented programming has always promised.

We actually decided that that basic COM approach we would not restrict to be a Windows-only thing. And so working together with a number of partners, we created the COM libraries for Unix as well.

Now, some of the capabilities, the rich applications we write that use COM we don't provide over it. But anybody who's developing in COM, they can take that capability and run it in Unix as well.

And so it's all part of sort of the embrace-and-extend strategy, where COM applications can run on Unix. You can use our services for Unix to have the coexistence, while, you know, what we view as a major trend is more and more of those users will be based on Windows NT.

And so a lot of applications vendors now say, okay, we have both versions. But the one on Windows has this tighter integration capability. And so even beyond the price-performance, it might be preferable there.

Where this COM -- how much COM richness we provide on Unix is something that's driven by ISV demand. So I hope we're reaching your requirements. I would love to talk about how we're doing.

ANAND CHANDRASEKHER: Question from this side of the room.

Q. From Audience: I'd like to hear from both of you, what role is LINUX playing in this growing and competitive workstation market?

BILL GATES: Well, it's simple. (Laughter.)

BILL GATES: We compete with many operating systems. LINUX is one of the ones that we compete with.

And we are -- we're not seeing that many people move over to LINUX. In the university environment, it's been very popular.

But in terms of bringing all the elements together, testing all the pieces that go into it, it's an incredible effort. Windows 2000, by the time it's shipped, will have over a billion dollars of engineering development into that. And a huge part of that is testing all of those pieces, making sure they come together the right way. There's over a thousand people who are dedicated to the testing there.

So in some ways, you're asking the most biased person you could possibly ask. (Laughter.)

CRAIG BARRETT: I'm not nearly as biased as he is on this topic. (Laughter.)

CRAIG BARRETT: You notice when I pointed out with the IA-64 family, we already have eight OSs booted up on the Merced simulator. But, strictly speaking, the great numerical superiority demand is really IA/NT, although we are supporting all the Unix or LINUX versions as well.

ANAND CHANDRASEKHER: Question from this side of the room.

Q. From Audience: Thanks. Question for Bill.

I guess first, how much of an advantage -- well, can you lay out, for example, how much you think that this transformation -- I notice you conveniently included a white paper on how to switch over from Unix to NT. Can you talk about the integration for the migration and Solaris having a 64-bit now or Sun with Solaris having a 64-bit maybe is taking away a little bit of thunder for NT when it rolls out.

BILL GATES: Well, as far as support for Merced goes, we will be there on day one. We have a great partnership with Intel on that, building the compilers, providing the feedback on the performance. And we're super happy with how that's coming together.

We look forward to the Merced getting done. And, you know, as they get beta chips out there, we'll have the beta software that goes with that. As they go final, we'll have Windows 2000, not even today's Windows, but the future one, that's available.

So we couldn't be doing more or getting better support from Intel on that.

In terms of what's happening on the desktop, there's a lot of market data people who are much better at tracking that than I might be.

There's no doubt there's a -- the desktop, a lot of people are seeing the benefits of the approach we're talking about here.

You see it as it comes in cycles. You know, for example, when Intel puts together a team to do a new chip, that's a milestone where they have a chance to bring in new tools. And that gives them a chance to use the superior platform.

We see that in lots of companies, whether it's the telecommunications companies, a lot of them have a plan over the next two years. They will literally move to have Windows on all their engineers' desktops.

So the trend is a very strong trend. We're very pleased with the support we're getting from people. And if you want numbers, I'm not the one for that.

CRAIG BARRETT: I showed the IDC slide, which shows the workstation conversion. And, obviously, the historic RISC offerings have a negative growth rate.

I think one of the nice advantages going forward to the 64-bit level is, you'll have a seamless architecture, basically, from top to bottom in terms of continuity and cost of ownership and maintainability.

I think people will find the same advantage there that we're talking about with just converting the engineering workstation people over to the same thing the rest of the desktop users are doing today.

ANAND CHANDRASEKHER: Question from this side of the room.

Q. From Audience: This question is for both gentlemen.

What's the single bold-point message you'd like to get out to design engineers?

CRAIG BARRETT: IA/NT. (Laughter.)

BILL GATES: Yeah. And it's a two-way dialogue. We're very interested in hearing, you know, what -- we're putting huge investments to make IA/NT attractive for these applications. And if there's anything we're missing in this, this is a great opportunity for us to hear about that and make sure we get 100 percent of the way to what you need.

ANAND CHANDRASEKHER: Question from this side.

Q. From Audience: Hi. This is for both people.

I noticed that one of the goals in Windows 2000 was to reduce the number of reboots.

And I was wondering, it seems to have worked. But when I rebooted, it seemed to take longer to actually boot up.

And I was wondering, how long in the future do you see before there's an instant-on technology? I know they have hibernation now where you can shut down and come back where you were. But I wonder if with the 64-bit, if they'll boot faster and allow the technical people to work more, be more productive.

BILL GATES: Well, there's several good points there.

Reducing boot time is something that, within Microsoft, I'm known as a real madman about that.

Windows 2000, by the time it ships, will have a lower boot time than Windows NT has had in the past.

As you've noted, in beta 2, the boot time was quite a bit longer. In beta 3, it was a little bit longer. And there's some tuning there. We will get those boot times down.

I'm using Windows 2000 on my laptop computer, and I basically never reboot. I go into suspend. I go into hibernate. I don't find any reason to reboot, literally, for weeks at a time.

Now, if it's gone to the hibernate mode, it does have to go back out to the disk and get all that state into memory. So it takes about ten or 12 seconds. It puts a nice message up there when it's gone to the zero power green state.

So dealing with these things and getting the boot time to be almost instantaneous over time, that's something that we're very focused on.

There's even some hardware things that Intel's working on that help us do that. But we won't go backwards and we will go forwards.

CRAIG BARRETT: Yeah, the net result has to be to get these things like T.V. sets and that sleep mode where even when it's, quote, "off," it's still drawing a couple of watts, just like your T.V. set does for instant on.

You see desktops which are out with that capability today. And you're going to see more and more by the end of this year.

And removing some of the legacy aspects simplifies the whole hardware/software combination.

This has to just get better and better and better going forward.

ANAND CHANDRASEKHER: We have time for one more question. I'll take it from this side of the room.

Q. From Audience: This is for both of you.

In your vision of convergence, Craig, and also the vision of the whole world becomes a plug-in to Internet Explorer, the interconnection bandwidth is becoming more and more a key point of the responsiveness of how all of this is going to happen. And I'd like to hear both of you on your visions as to so how are we addressing those interconnection bandwidth issues.

CRAIG BARRETT: I presume you're talking mostly about outside of the big business environment. Is that correct?

Q. From Audience: It'll be both. Both intranet and Internet, both inside and outside. The consumption of bandwidth that we're experiencing just in our company, and everybody else I talk to, bandwidth issues are becoming the big driver of responsiveness of all of the interconnections to your servers internally and to the rest of the world.

CRAIG BARRETT: Frankly, my biggest concern is how I started to answer the question, which is the remote worker, which someone asked about earlier, which is at home, and your bandwidth back in to the enterprise. I think we can scale from one- to a hundred-gigabit capability and sufficient bandwidth inside the environment with the existing technology. It's the regulatory issues and the capital burden issues or capital barrier issues of getting high bandwidth established to the small business and to the home worker that I think is the biggest challenge, not only here in the U.S., but everywhere I go around the world.

Frankly, I think some of the regulatory barriers there are bigger than the capital or technological barriers. Whether it's ADSL or getting cable or decent bandwidth to the home or small business.

BILL GATES: Yeah, I totally agree with that.

One other aspect of this is, as you unleash inside a corporation things like NetMeeting and really let people do the video sharing, it's very key for the network manager to be able to have policy management that says, "Which of these traffic streams, you know, the SAP transactions versus those video streams, get priority where there are bond links?"

And that's our case, you've got quality of service capabilities are dramatically richer in Windows 2000. It's another case where we've got Intel/Cisco on our side. It's really looking at this QoS capability. Because we want to make sure that the bandwidth things don't throttle the ability to show through rich software and rich microprocessor.

So I'd say on the enterprise side, I'm very pleased with that QoS foundation that is laid there. I'm pleased with the gigabit moves that are taking place there and the optic fiber that's being installed to connect all the locations. So I think we're in great shape on that part.

And for the consumer or small business, we are somewhat dependent on the regulatory environment.

ANAND CHANDRASEKHER: That'll be all. We're going to take a break now. (10:05 a.m.)

About Intel
Intel (NASDAQ: INTC), the world leader in silicon innovation, develops technologies, products and initiatives to continually advance how people work and live. Additional information about Intel is available at www.intel.com/pressroom and blogs.intel.com.

* Other names and brands may be claimed as the property of others.