Intel Developer Forum, Spring 2001

Craig Barrett
San Jose, Calif., USA
February 27, 2001

ANNOUNCER: Ladies and gentlemen, please welcome vice president and chief technology officer, Intel Architecture Group, Pat Gelsinger.

(Applause.)

PATRICK GELSINGER: Thank you very much. It's great to be with you again. Welcome to the eighth of our twice yearly Intel Developer Forum. It's a great privilege and honor to be with you again. As we kick off our developer conference.

Who is here this week? Well, we're somewhere right around 5,000 people for the week. As you can see, we have a large number of press and analysts, we have seven continents. Unfortunately we're missing Antarctica. We have a heavy contingent of the chief technologists of the industry and R & D. We're delighted and more important than anything is we always viewed IDF as to be the place the industry meets. And with 34% of the tracks and labs being run by industry players other than Intel, we're very delighted with the growing and expanding industry participation in IDF. And this is our trend and commitment for the future.

The depth of IDF continues to grow. Each time we get together we find new and exciting things to talk about. We have new areas. You'll see the first ever demonstration of InfiniBand during the conference. We'll talk about the new architectures that we're releasing. We're going to be talking about new programs that we're driving into the industry, such as our developer vehicle program that we're going to be launching this week. New networking, new server technologies, and of course at the heart of it is the leading technology training that occurs anywhere in the industry.

When we started IDF, we were largely a client-centric activity. Over time we've added the focus on servers. We've added increasing amounts of software content and finally an expansion into communications. Not only the depth of the conference but the breadth of the conference continues to grow each time we get together.

Each morning we're going to be kicking off the conference with keynote speeches that we believe will strike and communicate key trends and directions in the industry. This morning, I'm very delighted, we're going to have Craig Barrett and Paul Otellini today. Tomorrow, we'll be having discussion on communications by Ron Smith and Mark Christensen, and I'll be following that with a keynote on peer-to-peer. And finally on Thursday we have the directions for the client for the next ten years by Louis Burns and the server architectural directions by Mike Fister.

Once we finish IDF here, we begin the IDF world tour. As you can see by the graph here, following the U.S. events we go to Japan, then Taiwan, PRC, and Europe. Between the four international events we expect that we'll more than double the overall participation in the IDF conference of the spring.

It's amazing but to put on a conference like this, today as we're conducting the spring IDF, we're already well underway in planning the fall developer forum. We invite you, encourage you, implore you to join with us at participating at making this the geekfest for the industry, the premier developer's conference in the world.

We encourage you to invite -- to join as a sponsor, gold or silver, to participate in the showcase. If you haven't been over to the showcase yet, please take advantage of it. It's really stunning the amounts of industry players and demonstrations, new technology that's being shown there. Marketing opportunities.

We have some 30 third-party announcements occurring this week at IDF, and we hope that continues to expand as we go forward. And finally, to play a role as a technology leader as a lab or lecture. And of course join us in the international world tour of the Developer Forum.

More than anything I want to thank you for your participation. This is a great conference, a great week. We want to have great activities and the training that goes on, and we want to have a little bit of fun as well.

It's now my privilege to introduce our first keynote speaker of the morning, and you know, you've probably heard of Craig Barrett in a variety of ways, but this morning I particularly want to talk about Craig as the Intel silicon cowboy.

Now, you might not have known this but Craig is a rancher. He has a big ranch in Montana. And just to make my point, he has a bunch of horses there. Oh, I'm sorry, in Darby, Montana. So outside of Darby, Montana, he has a large ranch there and he has a whole bunch of horses, and three of the horses' names, just to make the point as an Intel silicon cowboy, one is named NASDAQ. How would you like to be a horse named NASDAQ? The second one if you want to guess is the Itanium Ingot, and the third one, anyone want to guess what it would be? Pentium Princess.

Somehow, I don't think my daughter would go for it if that was her horse's name.

So anyway, we have our Intel silicon cowboy running Intel Corporation, and normally when we introduce somebody as a keynote speaker, you guys are very generous and gracious and give them a nice round of applause. That's sort of a city slicker kind of thing to do so I don't want you to do that this morning. What I want you to do is give Craig a nice yee-ha when he comes out. Okay? Do you think you can do that for me? Let's just practice one time.

And by the way, this is a nice cowboy hat. Wrong attire for the event. Cowboy hat. And when Craig comes out, right, on three I want you to give a nice loud yee-ha; okay? Let's practice one time; okay? One, two, three: Yee-ha! That was pretty good, that was pretty good.

So ladies and gentlemen, my pleasure to introduce the president and CEO of Intel better known as the Intel silicon cowboy, Craig Barrett. One, two, three: Yee-ha.

(Applause.)

CRAIG BARRETT: What I do know is Pentium Prince is not going to be pleased to know that he's a princess!

(Laughter.)

CRAIG BARRETT: Of course, with a city boy like Pat, he could hardly be expected to know the difference.

(Laughter.)

CRAIG BARRETT: I want to, first of all, welcome you all to the developer's conference. My topic this morning will be to talk a little bit about Intel strategy and specifically talk about the architectures for the Internet.

If you look at Intel's strategy today, and we'll go through the presentation and describe it a bit, it really is that we're trying to develop a series of architectures which are suitable for the basic building blocks of the Internet, basic building blocks that you and the audience can build off of, innovate, bring great products into the marketplace in the business or the consumer marketplace.

So I'll try to concentrate on those architectures, describe them briefly, and then some of the follow-up keynote lectures and all the discussions later in the week will build off of those.

If you look at what's going on today in the industry, it's kind of exciting. It's a bit of a roller coaster ride. The NASDAQ goes up and down.

Andy Grove and I are having a good time. Unfortunately, if you see the little icon above, it really suggests we're on a downward trend at this point in time. But I've been in this industry for about 27 years, and we've been through a number of these cycles. And you always know that when you're going up, you're eventually going to come down or level offer, and when you're on the way down or level, you know that it's going to go up eventually.

And what we really have to do is prepare for the next upswing. And you really need to recognize that any industry slowdown that we have, whether it's localized here in the U.S. or it's worldwide in Asia, Europe, Japan, Latin America, those slowdowns are going to end, and you need to be prepared for the upswing.

The simple way to be prepared for the upswing is that you never save your way out of a downturn. You never save your way out of a recession. The only way you come out of a recession stronger than when you went into it is by creating great new technology, great new products. And that will be a bit about what the presentation is this morning. That is the theme for the entire session, which is how do we bring great new technology into the market, stimulate the market and bring our collective corporations, companies back on an upward track.

So the agenda I want to talk about today will be a bit about the environment. And we'll focus on the Internet and how vibrant the Internet is, what an early stage of the Internet build-out we have, which is why we should all be excited about the opportunity we have. I'll talk a bit about our four architectures for the Internet. And again, those will be amplified later in the week. And then I'll end by talking a little bit about silicon technology, which is obviously one of the core competencies that Intel has. And what I want to try to do with that brief description of silicon technology is assure you all that there's still a lot of life left in the standard CMOS transistor, a lot of life left in Moore's law, and you can look forward to many, many more generations of technology coming from Intel and other companies providing great technology, great innovation, great building blocks for products to bring into the marketplace.

So let's talk a little bit about the environment. And you can kind of pick your poison here, if you will. You can read about PC volumes or depending on who you read or what study you look at, the growth rate is slowing down. The PC is dead. Growth rates are not going to just slow down; there will be an absolute die-off of PC volumes. You can look at telecom equipment manufacturers reducing their orders and impacting the entire network industry. You can look at corporate IT spending. I'm really giving the total gloom and doom picture of the slowdown that's happening across the world, whether it's IT, whether it's handsets, a dot-com meltdown. You can pick any one of these and get totally depressed reading about it.

But I think there's a reason to be positive about all of this. There is absolutely a slowdown in the U.S. manufacturing world, and you can call it a slowdown or a recession. It hasn't spread to the rest of the world yet. It hasn't spread to the other markets. It hasn't spread to Latin America, Asia, Europe. It may not. It may be that the feds rate cut, maybe that Bush's tax cut proposal may be enough to have a soft landing on this. But the really important thing is to look forward into the future and see where we're going and what opportunity we have in the future. And that future is really the digital world.

The world is going digital. And whether it's communication, whether it's access to information, whether it's conducting commerce, we're all going digital. And this digital world is what is exciting and what is the opportunity for us all. And I think it's perhaps the most exciting time that we've had. Even though we're in kind of a slowdown today, if you look forward to the future, the build-out of the Internet, the build-out of this digital world is very much in its infancy. And there's a tremendous amount of infrastructure to put in place, a tremendous amount of people to reach, a tremendous number of corporations to become involved in this digital world, and I think that's what is exciting. And if you just start to do a quick tour around the world you can see some degree of what's happening in every one of the geographies with regard to this build-out of the digital world.

Look in Asia, the compound growth rate of Internet usage is about 45 percent a year growth. Forecasted to have something like 400 million users by the year 2004. Within a couple of years, 25 percent of all those people online will be in Asia. There's a huge infrastructure build-out to accomplish this. Whether the infrastructure is wire line, wireless, server farms, what have you, tremendous infrastructure build-out in Asia. Very hot spot in the world in terms of the market today.

If you look in Japan, equal excitement in Japan. PC users and Internet users are growing something like 62 percent a year in Japan. NTT Dokomo in their iMode, the digital handheld device, basically went from zero to 17 million users by the end of last year. Over 50 percent of the population are wireless subscribers. And Japan will obviously be the first to go to the third generation of wireless capability, have a high digital bandwidth to communicate with. Immense amount of excitement there.

If you continue around the world, Europe forecasted to have more Internet users than the U.S. in a couple of years. They have the highest mobile penetration in the world with over 50 percent.

You get to 70 percent Internet penetration in some of the Scandinavian countries like Norway and Sweden. And finally, something that most of us have been lobbying for an appreciable amount of time in Europe is, in fact, the price of Internet access falling dramatically. We've seen in Germany, for example, Internet access charge rates go down by 75 or 80 percent.

So whether you're in Asia, Japan, Europe, an immense interest in building out the Internet infrastructure. And if you look in the Americas, still the center of the Internet activity, still the center of Internet commerce, something like nearly $900 million of commerce done in the U.S. over the Internet last year. Very small fraction of it being the business-to-consumer type of commerce, which is where the real meltdown is occurring.

The real hot bed of activity in Internet commerce is and continues to be and will be for the foreseeable future the business-to-business commerce. That's where you're getting in complete renovation of the supply chain analysis, business-to-consumer, business-to-supplier, business-to-customer type of activities. It's this business-to-business deal which is in fact changing the way we conduct our daily transactions in our corporations.

And it's not just in North America. If you look in South America they're forecasted to be 50 million Internet users within two years.

So wherever you look around the world, we're seeing this Internet build-out. Very much in its infancy, high growth rates, immense opportunity for all of us.

So I'm excited about the Internet being the engine for growth for the future. This build-out is just started, so it's a huge opportunity for all of us. And what we have to do is worry about developing for this Internet build-out and developing the great products for this concept of the digital world. The digital world, whether it be communications, information access, or commerce.

Those are the exciting aspects of this.

And what we have is the opportunity to stimulate this growth with great new products, to grow our way out of a recession with great new products. And every time we have a slowdown in this industry, you can be sure of a couple of things. One is that technology doesn't slow down. There will still be technology innovation. And those companies who are at the leading forefront of technology innovation will be the leaders as we grow out of a slowdown or grow out of a recession.

So it's new products and industry innovation which are important for all of us.

And what this requires is a commitment on our part, your part, everyone's part to invest for the future. You don't get great new technology, you don't get great new products without investing. And if you look at some of the things that Intel is doing in terms of increasing our R & D budget on a year-upon-year basis even in the face of a slowdown, increasing our manufacturing technology budget, both from a technology development and a capacity standpoint, our intention is to invest about $12 billion next year in these two items: Manufacturing capital and R & D for new product, new technology development.

This is what I think every one of us has to do, in fact; is concentrate in a slowdown in getting great new technology into the marketplace. This is the way we overcome a slowdown. This is the way we become more successful coming out of a slowdown than we were going into it.

And that really leads me into the next topic which is, all right, what are we doing as really a foundation for IDF and a foundation for you who are creating great new products for both business and consumers, what architectures are we bringing into the marketplace? And we have four architectures which I want to talk briefly about, and those four architectures really deal with the two basic competencies that Intel brings to the game. One thing we know a lot about is silicon, making integrated circuits. Complex integrated circuits. And the second thing we know a lot about is really how you handle digital information, computer architectures and how you take digital information from point A to point B, and when you get it to point B, what do you do with it.

So this concept of digital architecture and integrated circuits are really our two competencies. And what you get out of those are, in fact, four architectures which I think are really the foundation basic building blocks for much of the Internet build-out. And I'd like to go through those four with you briefly.

The first is really our core business which is the IA-32 or the basic core building block for the personal computer. It's exemplified today by the Pentium® 4 processor, it is exemplified by Microsoft Windows operating system, and it's exemplified by the concept of an Extended PC. And I want to say a bit about each one of those things, because I think they're all very three exciting things that you're going to get to deal with going forward.

Pentium 4 was specifically designed for the Internet, for rich information, whether it's audio, whether it's digital video, whether it's communications, whether it's imaging, whether it's entertainment. The product does exceptionally well at handling that type of data, that type of information.

This leads to the concept of rich user interface, and you get rich user interface really in two ways. One is by the operating system, and we'll talk about Windows XP in just a minute. You have a rich user interface, reliable, scalable, able to handle lots of rich information. And then you can add around that all sorts of devices, other digital consumer devices, to give the end user a great experience. And that's the concept of the Extended PC.

The extended PC really is that the PC is the center of your digital universe, and that what you can do is use that PC to extend your capability into the areas of audio, video, entertainment, communications, imaging. All of those rich user experiences which will drive people to use this new technology.

So we picked the Pentium 4, Windows, and really the next generation of Windows capability, which Jim Allchin will come out in a minute and talk to us about, and this Extended PC concept. A lot of excitement for the consumer. A lot of excitement for business going forward.

So let's consider this Extended PC concept. A Pentium 4 processor-based system with lots of other digital devices with a rich new user interface, a rich new capability for the consumer.

If you accept that model, then the PC is hardly dead. The PC is really the center of the digital world as far as the consumer is concerned. And everyone has been talking about this recently. At CES we talked about it, Bill Gates talked about it. Even our friend Steve Jobs at MacWorld talked about it about a week later. One of the few times we have Steve following our lead in something going to the marketplace.

But this concept of the PC being at the center of the digital world I think is key to all of us in terms of what we can innovate about, what we can bring excitement into the world about with your new technology and our new capability.

And perhaps I've talked enough about this. What I want to do now is have Jim Allchin who is the group V.P. at Microsoft in charge of platforms, and Jim is going to come out -- there he is.

JIM ALLCHIN: Hi, Craig.

(Applause.)

CRAIG BARRETT: Why don't you just tell the audience a little bit about Windows XP and what we can expect to see from it.

JIM ALLCHIN: Well, we're still in the middle of developing as you know. It's the next generation operating system. It's based on the Windows 2000 enhanced kernel, so you get a lot more reliability, with a different design center, a design center based on making the extended PC come to life. And the design center is very focused on what we call experiences.

CRAIG BARRETT: Experiences. What do you mean by "experiences"?

JIM ALLCHIN: Well, experiences meaning taking it step by step. Taking a user all the way through everything they want to do, whether it's dealing with other devices that are connected to the PC in their home or at work, out to services that exist on the Internet, and not leaving the user in some lurch when they're trying to do anything like a photo scenario where you're trying to get it off the camera, into the PC, manipulate it, print it, archive it, go out into the Internet and share it with your friends or family or print it on the Internet.

So a couple of weeks ago we did an unveiling before we did beta 2, and we talked about the experiences. We highlighted a few of them. We highlighted the photo, music, video, and remote assistance scenarios. But that was focused on simplicity, trying to make the point that my mom is going to love this Windows XP.

But today how about we show a little bit about power?

CRAIG BARRETT: Power. I like power, actually.

(Laughter.)

CRAIG BARRETT: Why don't you show us a little bit of power.

JIM ALLCHIN: Okay. Come on over here. We have a machine that's a 1.4 GHz Pentium 4, and I think it's got 256 megs of memory. Obviously this is a great machine. And hopefully, right now you can see on the screens this desktop. You should notice that it looks pretty simple. High color, but it should look very, very simple. A lot of absence of something, meaning the icons there. You can put icons there but we've made it so you can take the whole screen real estate and keep more things going on without having to constantly keeping Windows over so you can get to "My Documents" and the like. If you want to do that you could. But the point of this is it looks very simple.

Let me pull up the start menu. You can see it looks quite different than the typical start menu. You can see it's easy to see that Craig is logged in here and we have a set of frequently-used operations here, things like a task of getting on the Internet or e-mail.

I mentioned a little bit about these tasks. Let me just do a quick highlight here. This is a picture of my dining room table, it turns out. Here is some tasks here on the top left. It says view as a slide show, order prints from the Internet. And for any of the particular items that I select here, there are operations that I can perform on them.

There are different views that I could do here. Let me just go up and do thumbnails and pull up Andy's picture there.

CRAIG BARRETT: Andy? Andy?

JIM ALLCHIN: I'm sorry. Whoops.

(Laughter.)

CRAIG BARRETT: Remember who invited you down here today!

(Laughter.)

CRAIG BARRETT: Andy is that other guy, you know. He's the one who you're making it real simple so he can learn how to use a PC.

(Laughter.)

(Applause.)

JIM ALLCHIN: Oh, jeez.

CRAIG BARRETT: Going to send you back to Redmond, I think.

All right.

JIM ALLCHIN: Moving right along.

As you can see, if you have a folder, we can do a little template of what is inside the folder showing the pictures. It's not just tied to pictures. If I go back and I look at music, you can do the same thing. Here is a Santana album, again with the task highlighted there. Very, very simple to use.

It doesn't stop there. If we go into the control panel, you can see that we've made it much simpler. Pick a category, say I want to create a new user, and pull up this, and just say pick a task and have a new account created in a second.

But I said we were going to show power.

CRAIG BARRETT: I want power.

JIM ALLCHIN: So this is a -- something you should be familiar with. Turn up the audio.

CRAIG BARRETT: I've seen those guys before.

JIM ALLCHIN: So let me just show you something here. Did I say fast?

CRAIG BARRETT: You said fast.

JIM ALLCHIN: What I'm doing is I'm switching in and out of the welcome screen, which is the way you connect and log in. And as you can see, it shows how many programs are running in each of these. There's an account for Craig, an account for me. And I can go back in here and continue to play. It's actually at the beginning. I can blow this up and go full screen. Pull it back.

So let me leave that running and let me go into my account. As you can see, I just switched to a brand new account, and of course you can configure it any way you want. That was one of the problems before in a family PC or shared PC setting where it was very difficult in order to have your own settings of whether it be your home pages or documents or your music.

So to give you an idea, while that other video is playing, I have a command window here, and I'm going to start a compile going, which is from our SDK. And I'm going to start some more music. Turn up the volume here.

So we're playing music, it's not interrupted, we're doing a compile, we're doing video running in the other session. It's pretty impressive. But of course, everybody who is compiling code needs to be able to not only listen to music, they need to be able to play a game as well.

(Laughter.)

JIM ALLCHIN: So this is bringing up Tomb Raider.

CRAIG BARRETT: You do let your software jockeys play games at work?

JIM ALLCHIN: Of course.

(Laughter.)

CRAIG BARRETT: Our IT guys are a little more conservative than that, but okay.

(Laughter.)

JIM ALLCHIN: So when you stop and think about what we're showing here, we're showing quick, fast user switching between multiple sessions. We're showing lots of audio and video going on concurrently. We're showing a compile going on in the background with a much simpler graphical presentation. We think this shows a tremendous amount of power.

So we still have a little ways to go. We're just at beta 2. We have a few more experiences to lay out, but I think you will all agree that wouldn't you rather have this on your desktop? I know my mom is going to want it and I know our developers are going to want it, too.

Craig, thank you very much.

CRAIG BARRETT: Thanks, Joe.

(Laughter.)

(Applause.)

CRAIG BARRETT: We're really excited about Pentium 4-based machines and Windows XP and the extended concept. You're going to hear a lot more about that this week and going forward.

The second architecture that's important here is really something called the Intel ® Personal Internet Client Architecture. And again we provide basic building blocks here and think of the Personal Internet Client Architecture as being the architecture that's in handheld devices and cell phones.

You heard a lot in the press about mobile access to the Internet. You're hearing a lot about the combination of mobile access to your desktop PC. And this is what this architecture is really all about. It's about a different microprocessor than the IA-32 family, something called IntelXScale® microarchitecture, which is tremendously scalable all the way from a double "A" battery up to a gigahertz. You can scale the power to the needed level for the application. If you wanted extended battery life or strong power you get your choice of that range.

The beauty of this though is really extending the PC with the handheld device. And whether it's a cell phone or an iPAQ device, and I have a couple of them sitting over here. We're all familiar with the standard sort of Palm Pilot or iPAQ device. This one happens to be equipped with an 802.11 wireless interface. What you can do with a device of this sort is in fact pick up any application from a spreadsheet to a Word document to a streaming video off of your PC and play it on this handheld device. High data rate, something like 11 megabit.

But this is really the early stage of this type of PCA type of architecture. What people are going to do is really combine this handheld computing device with devices such as this which are, in fact, going to be the combination of GSM cell phones and handheld computing devices, something you could use as a GSM cell phone, and also have immediate access to all the information that you have on your PC.

I think that that's really the critical aspect here, is to not look at these type devices as alternatives to PCs or competitive to PCs in terms of Internet access, but really as adjuncts to PC.

The most popular handheld device today is a Palm Pilot. It's popular because it's an adjunct to the PC. It allows you to synchronize with your PC.

Wireless capability and also wireless Internet access capability really give you a further extension of your PC with your handheld device. We're very excited about this architecture. You're going to hear a lot more about this from Ron Smith later in this week. But it's this Personal Internet Client Architecture or PCA for short which is the second architecture which we're building building blocks, for which we're very excited about.

The third architecture is really the Internet Exchange Architecture. And this is an architecture which provides basic building blocks for networking equipment. And the way to think of this is not just that it's a processor. There is a network processor which is the center of this, but there are a whole lot of other things building up this architecture.

It is truly a top-to-bottom architecture. It's got the network processor, got application software, got an application processor that sits on top, and then it has all the levels of interface that you need to get to the transport mechanism down to the physical layer devices where you're connected twisted copper, coax, fiber, whatever it might be.

So this Internet Exchange Architecture is really going to be the basic building block for network and telecom type companies.

We plan to introduce over 35 new products associated with this architecture family through the year, all the way from physical interface devices on up to variants of the network processor.

And we could show you lots of demos. What I want to do is just for a minute show you a couple of examples of how companies are using parts of this architecture. And again, it's a total architecture, hardware/software tools capability, that come to play.

And I want Anthony to come out. We're going to show you a couple OEM systems built on this architecture.

Why don't you tell the audience what we have here.

ANTHONY: Craig, the first system we have is the 41554 optical platform. This is designed to transmit from DS 3 speeds up to OC 48. And it provides a good way for customers to move voice, data, and video traffic.

CRAIG BARRETT: What sort of advantage does Cisco get from using this Internet Exchange Architecture?

ANTHONY: You really get two advantages. One is a time-to-market advantage. When you look at Intel Exchange Architecture, we design our components, our software, and our tools to work together to get our customers into the market more quickly.

CRAIG BARRETT: What else have you got here for us?

ANTHONY: We also have a Fast Ethernet switch from Nortel Net Gear. This is the SF 554, and this gives us an opportunity --

CRAIG BARRETT: It's a Motorola device that doesn't work. (Laughter.)

ANTHONY: Clearly the mike is not IXA enabled. The net gear device is a Fast Ethernet switch for 24 ports and what that does is allows PCs and workstations in a corporate environment to talk back and forth to each other, and also get out towards Internet access.

CRAIG BARRETT: Great. And what I think you're going to see from this architectural is a whole array of new products coming out from the OEM community later this year. Again, we're going to provide at least 35 new building blocks as part of this Internet Exchange Architecture capability, but the time to market, effectively the standardization, the reusability of software code building off this architecture I think is what is going to make it very, very popular.

ANTHONY: Exactly.

CRAIG BARRETT: Thanks.

ANTHONY: Thank you.

(Applause.)

CRAIG BARRETT: The fourth architecture obviously is the server architecture. And we have really two server architectures. One is the high performance IA-32 family, Pentium® III-based, Pentium® III Xeon™-based servers, and that will move forward with Pentium 4-based servers a little bit later this year, and you'll hear more about this from Mike Fister later in the week.

But there's a very exciting aspect here which is the new architecture which is the IA-64 or the Itanium™-based processor architecture which is starting to be rolled out now. We have the first instantiation, the Merced. We'll have the second in production towards the end of this year.

What I want to do is give you a little bit of a demonstration of what you can do with the processing power that you see in Itanium processing based systems. And we're going to have Rob Pennington from the Center of Supercomputing Applications come out and show you what you can do with Itanium processor-based systems.

Rob, welcome.

ROB: Good morning, Craig.

(Applause.)

CRAIG BARRETT: Why don't you tell the audience, they may not be all familiar with NCSA. What is it and what do you do?

ROB: It's the national center for supercomputing applications. We're located at the University of Illinois. NCSA has been charged by the national science foundation with developing and deploying the infrastructure to support computational and scientist and information technology into the 21st century. We've been one of the pioneers of Internet technology for quite some time with the development of the first graphical browser.

In addition, we're following this on with a peer-to-peer networking technology referred to as a computational grid.

CRAIG BARRETT: So basically you're going to give any scientist access to this immense amount of computational capability?

ROB: Basically you plug into the grid and you have access to other systems.

CRAIG BARRETT: What have you been doing with the Itanium-based systems?

ROB: We've been working with Itanium for about a year now. A lot of this has been in porting and tuning high performance applications to the Itanium-based systems. But more importantly, maybe, is we've been working in the software infrastructure to use these systems as part of large-scale clusters.

So with a large-scale cluster, with this clustering software using Itanium-based systems, we can look at some of the fundamental questions in science.

CRAIG BARRETT: I'm interested in the fundamental questions of science, too. Especially interested in what makes the NASDAQ go up and down which is a fundamental question of science.

ROB: Can we do something easier?

CRAIG BARRETT: All right, let's do something easier than that.

ROB: So what we have here is a cluster of four Itanium systems. These are four dual processor systems running a commodity interconnect between them. These particular systems are running 64-bit Windows right now. And we have the software infrastructure on these systems to be able to use all eight processors on a single application.

CRAIG BARRETT: So how much computational power do you get out of these eight processors?

ROB: Well, the peak power on individual processors is about 3.2 gigaflops. This rack of four is about 25 gigaflops.

CRAIG BARRETT: Pretty good computer, in other words.

ROB: This is not bad. The Itanium architecture is becoming competitive with traditional supercomputers.

CRAIG BARRETT: Show us what it can do.

ROB: This is an application we can run on the architecture, this is the simplified piece-wise parabolic method. This was developed at the University of Minnesota and it's a gas-flow dynamics problem. This will run across all eight processors and you'll see the performance monitor go up as the systems come on line and begin to run.

What this is doing is calculating the impact of a mach 7 shock on two concentric gas bubbles in an ambient medium. And right there you see the effects. And the thing to take away is that the shock wins in this case.

CRAIG BARRETT: The shock wins.

ROB: The shock definitely wins.

CRAIG BARRETT: All right. Considering we had some music playing with some of the background stuff before, probably some Jimmy Hendrix would be suitable with this.

(Laughter.)

ROB: The tie-dye colors are nice and they're significant because they tell you the density of the systems.

CRAIG BARRETT: Truth be known, what we're doing is basically it takes some period of time to simulate this complex calculation.

ROB: In fact, this particular calculation takes about 25 minutes on the Itanium cluster. And this is pretty significant. This is competitive with traditional supercomputers.

CRAIG BARRETT: What does it take to do it real time? Rob if I had four or five more racks sitting up here, if you could see your way through to sending them our way, I'd appreciate it.

CRAIG BARRETT: I'm talk to you later.

(Laughter.)

CRAIG BARRETT: What are the future plans?

ROB: Well, we have had some very good experiences with clusters and Itanium, and this year we're going to be putting in the largest Itanium cluster that we know of. This will be a teraflop system. This will be used by computational researchers to do research in nano materials up to stellar evolution and galaxy formation. Also will train the next generation of computational scientists. But one of the interesting things is you can also use it for energy generation, things like laser fusion.

CRAIG BARRETT: Governor Davis probably would like to hear something about that to solve the electricity problem here.

Thanks, Rob.

ROB: Thank you

(Applause.)

CRAIG BARRETT: These are the four architectures really for the Internet, the basic building blocks. And you can look at them in a variety of ways. You can look at them basically we bring competencies of silicon and architectural design to the IA-32, the IA-64, the Itanium processor-based family, the Personal Internet Client Architecture and the Intel Exchange Architecture. You can look at those four building blocks really as separate and distinct but it's probably more interesting to look at them really as part of the next generation network that we all know is going to happen.

We all know we're going to have a common network, common communication protocols. We're going to go digital from end-to-end. It's going to be packet-based and not switches-based.

So we're looking at clients attached to one end of this next generation network, and these can be PC or wireless handheld type clients. We're looking at the network infrastructure, and then we're looking at the enterprise at the other end. And these four architectures are really the basic building blocks for this next generation of network.

And you'll hear a lot more about this going forward this week and in the future from Intel and what these basic building blocks can do.

The concept here, though, is it's great new technology, and it's new products, new technology that lead the way to success. And when you start to talk about new products and new technology, we automatically come up and start to talk about, well, what are the basic fundamental technology and technology limitations. And I want to spend just a minute and talk about silicon and silicon technology, because that's one of the things that we do, and not one of the things we often talk about anymore. We tend to talk about processor performance or bandwidth or things of that sort. But fundamentally, this all goes back to basic silicon technology.

So I want to talk about silicon for the Internet for a minute, and talk first of all about size is important. And in terms of size, there are two aspects of size, big and little here. And the littler the size of the transistors that you manufacture and the big is the size of the wafers that you build these transistors on.

I want to give you an idea of what we're talking about. This is kind of state of the art today. These are eight-inch wafers and these are what most modern manufacturing plants are full of and this is what we make Pentium 4 processors on and Itanium processors. They're kind of passe.

This is what we're going into production on this year which are the 12-inch wafers, medium sized pizzas. We'll get to the large and extra large pizzas in several years from now when we go from 12-inch wafers to probably 16 or 20 inch wafers. But this is what the industry is moving towards.

What I want to do is just give you kind of a little bit of a walk-through in terms of the size effects that we have between the size of this wafer and the size of the smallest dimensions that we can print on it. And the way I want to do that is kind of give you a walk-through study from a wafer down to the smallest dimension on the wafer.

So to put this in perspective, I want you to think of this as the size of the Earth, basically, 8,000 mine diameter. And then we'll look at the small dimensions. I don't want to talk so much about nanometers and microns and angstroms because you aren't too familiar -- some of you aren't too familiar with those but I'll talk about things relative to the size of the Earth.

12-inch wafer equals the Earth, 8,000 miles in diameter.

I won't throw it. They were nervous I was going to throw that all week.

And we'll want to do this in terms of state of the art manufacturing technology. 12-inch wafer, we're going to talk about it in terms of a .13 micron technology with copper metallization, products that run in excess of 2 GHz, that is 2 GHz microprocessors, transistors run in excess of 20 GHz. So obviously individual transistor speed is much larger than the device that the transistor makes up.

So 300 millimeter wafers, 12-inch wafers, .13 micron transistors. And what we're going to do is take you through a magnification journey of this as fast as we can.

We'll start off with the wafer and then we'll zoom in from the wafer down to a specific die or one of those little square hunks of silicon on this test die. And what you're going to see through a range of optical electron, both scanning and electron microscopy and lattice imaging microscopy techniques down to the finest dimensions here.

So let's take this journey from the wafer down to a test die. And the test die we're going to come through here is in fact a test die on this device. And this happens to be something which is a little bit less than one square inch on a side. It's 22 millimeters on a side. The lower portion of the device, which is all that regular pattern, happens to be a bunch of SRAM. The upper portion happens to be a bunch of decoding logic and control logic for this test device.

There are only 160 million transistors on this device. If you prorate that to the wafer as a whole, we're printing something close to 20 billion transistors on that wafer.

So this is the first step. And in fact, if that wafer was 8,000 miles, this test die is only about 580 miles on a side. A little bit shorter than the length of California.

If you want to look at the metallization that's on that, what we're going to do here is zoom down into the metal structure, and this is a copper metal structure and it's called the -- the process is called damasking after Damascus rug-making, or, in fact, there's some degree of similarity where you print these copper lines into spaces or you really electroplate the copper lines on the device similar to the rug making techniques which is why they call it damasking. You don't have to remember that. Forget it.

But this shows the copper metallization structure, six layers of metal. What we've done is etched away all the interlayer dielectric so you're seeing these freestanding copper wires, six layers looking down on the device. Again, this will be on a .13 micron technology. And that's kind of interesting but not so interesting because we're looking down through this lattice work of six layers of metal. It's much more interesting to get down and look at the actual device or a cross-section of the metal.

And what we've done here is taken basically an actual cross-section and then kind of used the computer to make a three dimensional cross-section out of it showing the six layers of metal down to the little tungsten plugs. The tungsten plugs are those white little plugs which actually make electrical contact with the active silicon layer; that is, the sources and drains of the transistors or the switches here.

So again, if you look at this in terms of dimensions, this thing is about a thousand feet high relative to that 8,000 mile wide wafer that we talked about before. So it's -- On the Earth, functionally this metal stack is about the same height as the Empire State Building, if you put it in that perspective.

Now, what we need to do is to go further down into this device, and we'll go down into what the fusions look like on the silicon device. And basically what we're looking at now is we've taken rid of all the metal and we're actually looking at the diffusions on the active silicon surface, the sources and drains where the gates go over, et cetera, et cetera.

Effective magnification here relative, again, to that wafer the size of the Earth is about 250,000 X showing these. But that's not what's really interesting. What's really interesting is to get down and look at the transistor itself. And we're just going to zoom in on that transistor. And this is what it looks like. For a .13 micron transistor, the actual transistor dimension is about half of that. It's 70 nanometers, or .07 microns. It's one of the games that this industry plays in terms of dimensions, but .13 micron actually gives you transistors about half that size.

Let's look at that transistor. It is truly a beautiful thing when you look at it.

(Laughter.)

CRAIG BARRETT: What we have here is this little stack with the snow peak on top is, in fact, the polysilicon transistor with a little silicide on top for connectivity, and there's some nitride and oxide spacers on the side. The interesting dimension is in fact, that polysilicon stack inside the poly nitride and oxide spacers. That's .07 microns.

That dimension on the face of the Earth would be about nine feet wide. So we've gone from the dimension of the Earth now to something that's nine feet wide, which is the size of that relative transistor.

But that's pretty cool, but what's really cool is when you start to look at not so much just the spacing of the width of that transistor but you look at the dielectric underneath that gate between the poly gate and the active silicon surface. That gets to be really pretty small. And what I want to do is show you what that looks like. And you can use lattice imaging techniques and electron microscopy where you can, in fact, image the actual atoms. And what I've shown in this slide is, in fact, that.

On the lower right-hand side of the screen you're looking at really we're imaging the individual silicon atoms in the silicon substrate. The top left hand we're imaging the silicon atoms and the polysilicon gate and in between those two we have the gate oxide. And that gate oxide is 1.5 nanometers. You can translate, that's about five atom widths. So that's silicon dioxide with about five or so atoms or 15 angstroms, each atom being about three angstroms.

That dimension is about two inches relative to the diameter of the Earth or that big wafer that we talked about before.

That's kind of where we're going, or what we have today. If you want to look at where we're going, we always have about two generations of research we're working on ahead of us. So we're about to put .13 micron into production. We're also already doing research on .1 micron and .07 micron. But I think it's kind of interesting to see the smallest transistor that we've been able to make yet, not just a single transistor but an array of these transistors that work. And that's a 30 nanometer or .03 micron transistor. And what we've shown here is the size of that transistor on the left compared to a DNA strand on the right. And those two gold-decorated DNA strands on the right give you a relative order of magnitude of the size of the transistor relative to them. In fact, that large dark particle on the leftmost DNA strand is a ten nanometer gold particle. These are gold-shadowed techniques to illuminate these DNA strands.

We've already, then, been able to make something which is .03 microns, which is about three generations from where we were today.

The message I want to leave you with is there's a lot of leg left in the silicon technology: Five generations or so yet to go.

You can imagine the compute power that you'll be able to put into one of these systems, and you can use something like the .03 micron transistor to cram maybe 400 million transistors into a microprocessor that's going to run at 10, 20 or 30 gigahertz. You'll be able to do demos of the type that Jim Allchin was showing us before, but rather than having a keyboard and a mouse there, what we're going to be able to do is walk up to that computer and say, gee, I've got a video file here, somewhere stored, from the fishing trip I just did in New Zealand. I want to show my friends the brown trout I just caught, so all I do is walk up to the computer and say: Go to the video files, go to the New Zealand trip, find the picture of the brown trout. And we'll get the streaming video back to us. Very exciting.

Let me quickly summarize. I think the Internet is the growth engine going forward. It's in its infancy. We're just starting to build out. It is going to be our future.

We have a series of architectures for that Internet. You'll hear a lot about those architectures this week. Very exciting technology development in each of them. Lots of exciting building blocks that you can build on top of, create value, create great products, bring them to the consumer, bring them to the business world.

New products will lead the way out of a recession, or slowdown that we're in, and these companies that invest in these products will be the winners moving out.

And don't be afraid. We've got lots and lots of silicon technology yet to come. We have generations to come from where we are today, building off the same basic switch, the same CMOS transistor that we've come to grow, live with and love over the last few years.

What I want to do now is move on to the next presentation, and that's going to be Paul Otellini who is our executive vice president in charge of all of our Intel architecture products. And Paul is going to come out and talk about really platforms, compute platforms of the future and talk about a couple of the architectures I have mentioned this morning, the IA-32 and IA-64 architectures.

Paul.

(Applause.)

To Paul Otellini's IDF Spring 2001 Keynote Speech

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