Intel Developer Forum, Spring 2001

Paul Otellini
San Jose, Calif., USA
February 27, 2001

PAUL OTELLINI: Thank you, Craig. And good morning. As Craig said, what I'd like to do this morning is drill down a bit into two of the architectures that Craig described, our 32 bit microprocessor architecture and the Itanium™ processor family, and tell you, as developers, both hardware and software, what our directions are on those two architectures for the next year, year and a half specifically, and what kinds of platforms we think you'll be able to enable and build and bring into the marketplace as a result of them.

At the highest level, I think what we're talking about this year is really taking segmentation of our product line, of Intel's product line, to the next level.

As you know, we've been segmenting our product line for about five or six years now by making increasingly specific microprocessors and chipsets and platforms to deal with the different needs of different levels of computing.

And we're doing this in response to customers' requirements. As computing matures, people are more demanding, more specific about what they need in their type of computing.

And it's very interesting, because these platforms are coming to us at a time, I think, when we are about to see the next wave of computing. We're sitting here at the beginning of a change, a fundamental change in the usage of computers that is as significant of a change as we've seen over the last ten years in the way people interact with machines.

And just like we've seen with every trend in computing in the last 10 to 15 years, consumers are leading the way. They lead the way in terms of Internet use. Yes, business is moving fast and yes, we expect transactions in e-Commerce to overtake consumer use on the Web. But consumer patterns today represent 50 percent of Internet use.

Consumers are also doing other things. There're footprints in the snow in terms of what they intend to do with their digital world.

A number of new devices have been invented and consumers are adopting those. If you look at MP3 players, they've grown by a factor of four in the last two years, and are doubling again. Digital cameras are doubling every two years in terms of the amount of people. Digital video recorders represent two-thirds of the market growing at a 25 percent annual growth rate. Instant messaging now has 50 million unique users around the world. And Internet talk has moved from 300 billion minutes this year to three trillion minutes in the next four years.

All of these things are coming together to create a digital spiral. Everything that consumers do and care about in their personal lives is becoming digital. The way their entertainment is being digitized, their memories of their loved ones and their vacations and so forth are becoming digitized, the way they communicate with friends and families around the world is becoming digitized.

All of this is driving a digital spiral of the way we live in the world.

From our perspective, this opens up a threshold that Craig described in terms of the next wave of computing. The thing that Intel calls the extended PC. And it really represents, from my perspective, the fourth wave. The first wave, in terms of computing, was the thing that we all became most familiar with:productivitywith: productivity applications moved from discrete devices like typewriters and word processors and calculators to things that you could do on the computer more effectively and more efficiently.

Multimedia came on and allowed us to bring on rich data types. The Internet exploded and brought the whole world to our desktop.

And what we're seeing now with the extended PC is the ability to start integrating all of these digital devices into the one fundamental device that remains constant and evolves throughout that time frame, and that's the PC.

The PC really is uniquely positioned to be the device that manages all of that data, that archives all that data, that allows you to edit all that data, and most importantly, allows you to share it with friends or co-workers around the world.

And rather than just talk about this use pattern, I thought I would bring out Brad from Intel and show you some of the new uses that are emerging in terms of Pentium 4-based processor technologies.

Brad, good morning.

BRAD: Good morning, Paul.

PAUL OTELLINI: What have you got to show us today?

BRAD: A way to use the Pentium 4 to speed up a video encoding application. As you know, video encoding applications are typically very processor intensive and with the new Net Burst™ microarchitecture, the Pentium 4 and Flask, it's actually quite easy to take a very high quality MPEG2 video clip and convert it into a new start, MPEG 4.

What is going on here is we have a MPEG2 video clip and that would take quite a bit of time to download over standard lines and it is recoding it into MPEG4 format on the fly.

PAUL OTELLINI: You have something where you can actually do it over a regular phone line?

BRAD: Absolutely.

PAUL OTELLINI: That seemed to go fast.

BRAD: That's correct, and because this small form factor PC is equipped with a 2 GHz air-cooled processor.

PAUL OTELLINI: 2 GHz Pentium 4?

BRAD: Right. And let me show you what the resulting video clip looks like because it looks quite good.

PAUL OTELLINI: I think we've seen that clip before today.

BRAD: Yes, exactly.

PAUL OTELLINI: That's very interesting. What other kinds of emerging usage models do we see happening here?

BRAD: Lets step over here. What we have here is an application by Skyline Software Systems called Skyline and what we can do with this application is zoom in and look throughout the planet at various areas of interest.

So for example, say you wanted to buy a house. Say perhaps you wanted to move down to Southern California.

PAUL OTELLINI: Not a chance.

(Laughter.)

BRAD: If you did want to move down there, these images come actually from satellite and topographical maps I should mention. What you can do with this application and zoom around and get a feel for the type of area you might want to move to, if you wanted to move down to Southern California.

Say you wanted to find out where the nearest tennis clubs were, parks, so on and so forth. This application makes use of the Pentium 4 processor and lets you get a good look.

PAUL OTELLINI: That's really spectacular. What else can you show us?

BRAD: I've got one more application. And this is one well suited for the information age. This application is by a company called SmartForce and what we'll be looking at here is a cross-section of one of the most complex systems we can probably imagine, which is the human body. Let me start this off here.

"...is a model of the human head and we are going to look right inside it and see how the sinuses and nasal areas are structured."

BRAD: Pretty cool, huh?

PAUL OTELLINI: I never really dug physiology in class.

BRAD: Well, me either, but let me tell you one of the neat things about this that I think is worth taking away, is these applications right here actually are very strongly dependent on the processor being used. So the files that are actually being used here to generate this content is ?? it's not very big at all. So this makes it very ideal as an application for long distance learning, for applications that have low bandwidth connections available to them.

PAUL OTELLINI: So back to Craig's earlier point of increasing the richness. You can use this type of technology with regular phone lines to be able to deliver very high quality, very rich training to people on a worldwide basis.

BRAD: Exactly.

PAUL OTELLINI: Thank you very much, Brad.

(Applause.)

PAUL OTELLINI: Now, to enable this ramp, Intel is going to be ramping the Pentium 4 as quickly as we know how. The chart that's up on the walls now shows the last four microarchitecture ramps that we brought into the marketplace, not in terms of the total volume, because it's easy to have a faster unit volume because the market has grown sequentially over the last four generations, but rather as a percent of the total output of our factories.

So you can see that the top line, which is the gold line, is the Pentium 4 and it's ramping much faster as a percent of our total output over its first four quarters of life than anything we've done before.

Our intention is to really do two things. First of all, we want to ramp quickly now based upon the existing Intel® 850 chipset which is based upon our DRAM. And then you'll hear plans in the various breakout sessions over the next two days about our plans for other chipsets, one in particular code named Brookdale, which will be out for the volume selling season this year that allows us to incorporate either SDRAM or DDR to be able to span all of the price points necessary to support that volume ramp on the right-hand side there.

Now, I talked about segmentation earlier and I talked about people wanting to get more out of their computers. There's more than one vector in terms of the richness and data types. That's the degree of get realtime connectivity that all of us need as users. You want computing to be where you are when you want it.

In mobile computing there's a very interesting trend going on. If you look at the shape of the mobile market from the '99 to '01 time frame, 60 percent of all the notebooks shipped in that three-year period are what the industry would call full size, classic inch and a half, two inch, full pound notebook that delivered the full feature set.

An interesting thing is developing in the '01 '02 time frame. 62 percent is going into the small and light, and many more are going into the mini notes and subnotebooks.

As a result of these trends, and we don't think they're diminishing. We believe the trend towards thin and light, ultra mobile will only continue over time. We have redefined what we are doing for mobile microprocessors and chipsets. And I like to think of this as optimizing our products around four different and distinct vectors in terms of what we want to deliver to notebook users worldwide.

The first, of course, is performance. Performance matters for everyone in terms of notebooks. People, first and foremost, want that desktop equivalent performance in their notebooks. And we continue to drive to deliver these.

We will have 1 GHz in production this quarter, and the first .13 micron processors out of Intel will be mobile microprocessors, and those will be delivered in the second half of this year. We've also, though, started thinking about things like helping to enable very compelling, ever lighter, ever thinner form factors. And to do this we've continuously lowered the voltage and power dissipation of our standard microprocessors to where they're running under a watt today, but also brought out in the last nine months a line of low-volt microprocessors that don't compromise performance and, in fact, today at the Developers Forum we're announcing a 700 MHz version of the Pentium® III low-power mobile product that actually will run down to 1.0 watts on average use in terms of battery optimized mode.

The third vector that we're optimizing along is battery life. All of us would like to have a longer battery life, and we are working towards nothing less than an all-day, eight to ten hour, kind of machine. Now, the microprocessor in the chipset is not the most significant power consumer in that equation, so we have to work with you, the developers in the industry, to optimize all of the other elements of the notebook manufacturing process and components to ensure that we can get the power requirements down and get that all-day life.

Intel is continuing to deliver new technologies here. Obviously the Intel SpeedStep™ technology, which allows us to run at higher frequencies under AC power versus battery power where you can save some battery life. But we've also recently introduced an ultra low voltage product line where products run under half a watt, and the fastest of those is now at 500 MHz. That was introduced in January in Japan, and you'll see us increment that product line over the course of the year.

So where are we going next after we optimize on those vectors? Well, the most obvious one is to integrate seamless connectivity into the notebooks. We are working with manufacturers to enable Bluetooth* and 802.11 to be incorporated seamlessly into notebooks this year, 2001.

As we go forward out of 2001, the first Pentium 4 notebook microprocessor will be delivered on our .13 micron process following the Pentium III based product coming out later this year. And as we look farther into the future into the '03 time frame, we are now developing a line of mobile optimized mobile properties that will integrate much more seamlessly all of these characteristics I've just described in terms of what notebook users demand.

To give you a better idea of what I'm talking about in terms of seamless connectivity and no compromised performance let me bring David out from Intel to give us a little tour of the notebook world.

DAVID: Hi, Paul. Did I sneak up on you? I've got a very mobile platform here coming through.

You know, you were mentioning the 700 MHz low-volt systems. I brought a couple of examples. This isn't one of them. But what I have on the left is a Hewlett-Packard OmniBook 700 MHz system.

PAUL OTELLINI: 700 MHz, low power?

DAVID: Exactly. And on the right I have a Toshiba Portege, and on these two systems I have a CPU intensive application, it's a music visualization software from Oozic.com. And what it is is a high-impact, immersive, interactive 3D experience that reacts to the beat of the music.

PAUL OTELLINI: So the bubbles are dancing to go the music?

DAVID: Exactly.

PAUL OTELLINI: Just what I need.

(Laughter.)

DAVID: Everybody needs one of those. So what we have is incredible performance. As you look at these, as you said earlier, when you take these and unplug them, you go on the road, these scale down the processor to 1.1 volts, allowing the system to run ?? or the processor to run at ?? or 1.0 watts.

But not only high performance and low battery life. There's also other systems that people look in for notebooks. And that is connectivity. Connectivity is what is empowering the users.

And also, different users have different form factor requirements. So I have an example of this, and that's why I brought this out.

And what this is is a 1 GHz, very sexy, dare I say, thin and light notebook from WinBook.

PAUL OTELLINI: 1 GHz? So what's it doing?

DAVID: Let me point out it's one-inch thick first. But on this system, I'm actually running a peer-to-peer application. I'm stream accompanying encoding live video out through a wireless LAN out to my other systems.

PAUL OTELLINI: You're capturing video life from this camera and streaming it out on 802.11?

DAVID: Correct.

PAUL OTELLINI: Where is it going?

DAVID: To all my friends around the world on this wall here.

(Laughter.)

PAUL OTELLINI: You have very few friends.

(Laughter.)

DAVID: Hopefully I'll have more after this speech. As you can see, let's start from the bottom here. Form factor, it's all about choice.

We have on the bottom here a subnotebook. It's IBM ThinkPad "i" series. It's ultra low volt notebooks running at 500 MHz and it's operating at less than one ?? less than half a watt, actually. I'm going to refresh that so you can see the streaming video.

PAUL OTELLINI: So it's receiving the stream wirelessly from this machine.

DAVID: Correct. So when that ?? as you can see, there's a slight delay in the video. But let's move on to our other 700 low-volt notebooks that we have. Of course we want to show that there's a lot of OEM support for this when it's being launched.

We have the Dell Latitude here as well as the IBM ThinkPad X20.

PAUL OTELLINI: And you're streaming all to them. There's one machine up here that seems to be broken or locked up or something.

DAVID: It's not broken or locked up. It's our super-secret high performance prototype that I want to show everybody. And of course, I mean, are you ready for this? This is the first public demonstration of Intel's first processor based on .13 micron technology.

PAUL OTELLINI: So this is that mobile Pentium III processor I talked about earlier?

DAVID: Yes, it is. And here we have it live.

PAUL OTELLINI: Is there a chipset that goes with this?

DAVID: It's actually using one of the 830M chipsets.

PAUL OTELLINI: How do we wake it up?

DAVID: I've added increased security to the system. What I have here is a Xyloc key from Ensure Technologies, and what it is is Bluetooth enabled so when I'm near the system it will come on.

PAUL OTELLINI: This is the same streaming video coming on there.

DAVID: Exactly. So the exciting thing about this key, though, is it's a Bluetooth device, and it's working in the 802.11B environment that is set up here on stage.

PAUL OTELLINI: So we avoid the conflict problems that one would normally see between the two technologies.

DAVID: Correct. Intel is driving solutions to have seamless coexistence between two different standards.

PAUL OTELLINI: Terrific. Thank you.

(Applause.)

PAUL OTELLINI: I think that was a very interesting demo that brings together a lot of those technologies in terms of optimizing the things that mobile users care about. We really want to continue to drive the seamless connectivity so that you can have the personal area network kinds of things, whether they're security-based or share-based, completely seamlessly interact with the wide-area network technologies like 802.11 so you can have this anytime connectivity that even Jim Allchin talked about in terms of the Microsoft Windows XP orientation as that drives into the marketplace.

The last area I want to talk about the enterprise or the server marketplace. This is where a lot of these technologies really come together. Not only are these required to connect individuals to others and individuals to businesses and businesses to businesses, but increasingly the enterprise and the server infrastructure is where the wired world and wireless world really come together over time.

Now, this server market segment has been growing very, very rapidly, something around 28 percent compound annual growth rate the last five years. And the chart you're looking at shows Intel's participation in the overall growth of the server marketplace from '96 through the end of '00. And we've done nicely in that time period. Our architecture now represents 85 percent of all the shipments of servers on a worldwide basis that occur.

The small green part on top of it are all the servers that are based on other than Intel architectures that are out there.

One interesting thing about this curve is you really don't see an Internet bubble in this. We may see some effect of that in 2001 but for the most part this reflects the overall build-out of the Internet infrastructure in the last few years.

In many ways this curve was built upon a technology that we brought out in 1999 that was what we called the Pentium® II Xeon™ processor and migrated that to Pentium III Xeon processor.

Xeon processors have had eight consecutive quarters of growth and in the last two and a half quarters we have refreshed that product line entirely to move to the large cache versions of the product line that bring higher performance capabilities to all levels of servers.

Now, one interesting metric on this is that Intel technology has traditionally been the best price/performance metric in terms of the marketplace, in terms of server marketplaces, whether you look on it on TPC-Cs or TPC-Ds, the price/performance has always been very good.

In the last year, based upon the new large cache versions of the Pentium III Xeon running at 700 MHz, Intel architecture now represents the top four TPC-C machines in the world in terms of absolute performance, and the highest one is a Compaq 192 processor machine that delivers 505,000 TPC-Cs at about half the price of its nearest RISC competitor which is the IBM machine on the bottom of this chart.

So we continue to lead in performance and we continue to lead in price/performance with this product line.

We're not going to stop, though. We're going to continue to evolve the 32-bit line, and there are a number of products, exciting products, coming into the market this year. The first thing you'll see it is an increase in the frequency of the Pentium III Xeon processor to 900 MHz to up about 93 percent from where it is today that will happen this quarter. And following that in the second quarter, the product that we had code named Foster, which if you think about it is a Pentium 4 microarchitecture-based Xeon, will come out. It will come out initially into low-end servers and workstations and move up the server food chain over the course of the year as we bring out beefier and beefier versions of that.

We would expect to bring that product out at 1.4 GHz and above and we believe that the performance and things like TPC-Cs will be about one and a half times the fastest Pentium III Xeon processors that we're shipping at the time.

So continue to refresh both the Pentium III and then the entire Xeon product line over the next four or five months.

Beyond that, in the Developers Forum in the next two days you'll be hearing about Intel's plans to bring out new and very exciting chipsets for the entry-level and mid-level server market segments, and really take advantage of the processor power that we're about to deliver.

And the last bullet on here is about Intel moving to deliver telco server building blocks. I talked earlier about the convergence of the wired and wireless worlds. A lot of that is converging on rack-mounted servers that we believehavebelieve have the potential to be standard Intel architecture-based machines that are NEBS compliant, that our computer M manufacturing clients with buy and add their building blocks and storage systems to and be able to go out and compete and win that business. So we're very excited about the developments in our 32-bit line.

The server market segmentation we've talked about in prior forms. I cut this a different way this year. Instead of looking at the price bands, we look at the wayness of the systems, the number of microprocessor per system, and how the market is segmented by server architecture types.

Clearly, most of the volume is still in the front end or the infrastructure servers. These are the one and two-way systems. They represent 84 percent of all the servers shipped last year.

Intel architecture represents about 89 percent of that market segment share today.

As we move into the mid-tier, the four-way and six-way machines, they represent about 12 percent of the market, and based on the strength of the Xeon market line we've grown to about 76 percent of the market share there.

The interesting thing is the back end, though, where the databases reside, all the billing systems reside, classically the land of the mainframes. These are machines eight-way or better. Those represent only four percent of the units in the marketplace, but they represent a much larger fraction of the hardware revenue in the server industry.

And on the basis of Xeon, we've moved nicely into that. We have about 37 percent of the market segment share today. But we want to move beyond that. We want to be able to address all of the performance requirements in the server market segments.

To do that we've been developing the Itanium processor family for five-plus years now. I wanttowant to give you a progress report on that product line.

Based upon the original Itanium processor, we now have 30 OEMs in the final stages moving towards production in the second quarter of this year. There are over 300 applications that have been optimized for Itanium, and hundreds more in development.

In October at our Exchange event, we demonstrated over 30 complete solution stacks running on Itanium. And I think one of the big value propositions for this product line continues to be that it represents the convergence of all the major operating systems in server space. Itanium ports are in very, very good shape for HPUX, for IBM's AIX, for Windows 2000 moving to XP technology, and for Linux.

So we have a number of customers rolling out pilots now in the hundreds with institutions around the world as they move into production over the next three or four months.

Looking forward, our enthusiasm about this market segment is undiminished. If we look at the Internet build-out as it happened, the number of people generating large storage capacity requirements, and the amount of security that needs to be enabled into each of these servers to be able to handle the demands, we believe that we have a product line and an architectural that is capable of addressing this in a very, very cost effective, high performance fashion.

To build out Itanium, we need to have a second generation. The first generation is shown on the left. This was the product that was code named Merced. It's a single core. It runs at about 733, 800 MHz this year. It's constructed with a back-side bus.

We've been in development for the last couple of years on the second generation product that is code named McKinley. McKinley adds a lot of value. It's clearly still very, very instruction-set compatible with Merced. All the software that was optimized for Merced will run on McKinley. But we've added more things to the processor. We've added more execution units. We've increased the bus bandwidth by a factor of three. We've integrated the Level 3 cache onto the die itself, and we've got very, very significant performance improvement as a result.

We are able to deliver with Itanium something in the range of 8X the SSL transaction performance versus an Ultra SPARC III 900 MHz machine, and something like 5X the business intelligence benchmarks on like DSS data mining than our fastest Pentium III Xeon processors.

We had recently taped out this product, and I'm very happy to report to you today that Intel has first silicon on the processor. This is a wafer of the first McKinley microprocessor, and this is the product itself in terms of the package it's going to be in.

To help share our enthusiasm on this product, I'd like to bring out Duane Zitzner. Duane is the president of the Hewlett-Packard Computer Systems Group.

Good morning, Duane. How are you?

(Applause.)

PAUL OTELLINI: So, Duane, are you excited about McKinley?

DUANE: Absolutely. This is outstanding.

PAUL OTELLINI: What do you think it gives to you?

DUANE: It's going to be great, looking at the multiple OSs we can run on it, the capabilities we have there. We're working with customers today on the current Itanium product, but we see the Itanium processor family to be crucial going forward.

PAUL OTELLINI: So you and I are clearly excited about this, but I don't know that the audience is terribly excited about seeing a wafer yet.

DUANE: How could they not be excited?

PAUL OTELLINI: Maybe they're not as silicon-centric as we are. Maybe we could bring out Steve Hunt to show us some real excitement.

STEVE: Good morning, Paul. It's not every day we get to take a new processor out for a maiden voyage demo run.

PAUL OTELLINI: In front of 5,000 people.

STEVE: That's right. So let's take a look. We've got three machines here running multiple OSs, and I think we'd like to start off with the one right here in the middle.

We've got here a McKinley system that's running the McKinley A0 silicon, and you can see it's running Windows operating system. This is the Windows XP. In fact, it's the same build that Jim Allchin showed over there. 64 bit Windows XP, beta 2.

What I'm going to do is go ahead and kick off a little rendering. We're going to take the McKinley floating-point units for a little test drive. And while I'm doing that, I'd like to just show ?? you can see the same user interface on the start menu. And here of course we've got the Internet Explorer, the entire functionality of Windows.

Let's go on ahead and minimize that and take a look at the rendering.

Now, what's really cool about using floating-point units and doing software rendering is we can do ray tracing which is a much more accurate way of doing the modeling than the traditional kind of throw a graphics cashed in there and use polygons. We're actually shining lights of ray ?? rays of light through the scene and reflecting them across all these materials.

PAUL OTELLINI: Can you show us what you mean by that?

STEVE: Sure. Let's look at other pictures that we've actually rendered here. There's a beautiful seen of a chess game and you can see here we have photo-realistic looking effects from a cinematic point of view, depth of field and refractions of light. There's a couple more here.

This is a little sunset scene. So let's take it back to this guy.

PAUL OTELLINI: You said we had other operating systems. What else does it do here?

STEVE: Well, yes. We've got multiple OSs, and I think that's one of the big strengths of the architecture, that it's able to run and support the kinds of mixed operating system environments that you might have in an enterprise computing world.

DUANE: Absolutely.

STEVE: Let's take it through a little end-to-end kind of concept. We're going to talk about a workstation. This is the McKinley as a workstation running the Linux operating system over here on the left; okay?

DUANE: This is all A0 stepping?

STEVE: Yes, running with all the features on pretty much nominal conditions we would use for initial bring up.

We're going to edit some content on the Linux machine and then transfer it over to another McKinley machine that's running HPUX operating system, and that HPUX operating system will have Apache Web server, and then we'll go and hit that Web site. So that's what we're going to do.

I've got a picture of the McKinley die and I'm going to dial down the opacity.

This is the ?? a tool called GIMP, Gnu Image Manipulation Program. So we're seeing now the McKinley peak. So we have a composite image of the McKinley die photo layered on top of the Mount McKinley.

I'll go ahead and flatten the layers into an image that we can use for the Web, and I'm go on and save it out. And I have a file name here that I'm going to use, Mac image.JPEG and I'll use the default settings you would use for an Internet image.

And what let's do is go on ahead and copy that image. I'm using the remote copy command. These machines are all networked together. So we'll use the remote copy command and send that image we just created on this Linux workstation over to the HPUX workstation.

And what we'll do is we'll pull up a remote console window on the HPUX McKinley system. Let's go on ahead, for those of you that know the famous U name command in UNIX, you can in fact see we are running HPUX on the IA-64 Itanium processor.

So let's go on ahead and set up a tail on the Apache file. You can see here the hits from this morning when we were testing it. Paul, if you would like to go on ahead and hit the Web server.

PAUL OTELLINI: You trust me to do this. I hit the Web server.

STEVE: There it is. You can see the content that's been streamed over over there on the other.

PAUL OTELLINI: There's the photograph.

STEVE: There's the image we just created.

PAUL OTELLINI: Hot off the Web. Thank you very much.

STEVE: Isn't this cool?

(Applause.)

STEVE: A0 silicon all running.

PAUL OTELLINI: Maybe a lot of you aren't familiar with silicon debug technology, but consider this is probably the world's most complex, sophisticated microprocessor. The two companies have had silicon now for about three weeks, and this level of functionality on an A0 stepping in our history is really unprecedented. This product is very, very solid and we're very excited about it.

(Applause.)

STEVE: Thanks a lot.

DUANE: Thanks much. Appreciate it. This is really excellent. Three OSs working great. Very, very committed to this program, so we're really excited about it. Great stability. Let's keep it going.

PAUL OTELLINI: There's a guy out there to take purchase orders from you.

DUANE: Yeah, yeah, sure. You always do.

PAUL OTELLINI: See you.

(Applause.)

PAUL OTELLINI: So summarizing on Itanium, we're very excited. We're moving towards platform release of the first Itanium-based platform systems next quarter. McKinley is in very, very good shape, as we indicated. We have sampled multiple computer manufacturers now. We're obviously in debug mode.

When we introduce that product, and we believe we're still on schedule to have pilot production late this year and production systems in '02, we're going to introduce not just the microprocessor but an entirely new platform. And Mike Fister in his keynote on Thursday morning will talk to you about some of the elements of that platform. But some of the key attributes of that will be bringing in the first versions of InfiniBand I/O and scalability port to be able to have a very high performance, seamless intersect connectivity for servers based upon McKinley.

So the summary that I'd like to leave you with today is that as Craig indicated, we have a very, very robust product line this year, and we are investing very heavily to be able to accelerate our leadership in each of the platform segments.

I think all of us collectively as industry developers have an opportunity this year to really take the slowdown in the overall level of businesses and use this as an opportunity to refresh and to renew our product line, to take advantage of all of these new building blocks from companies like Intel and Microsoft and Hewlett-Packard, to be able to build very, very compelling systems for end users.

We're excited about it. We think that the product line is better positioned today than it has been in years, and that the opportunity in front of us is very, very compelling.

Thank you, and enjoy the rest of the conference.

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