Intel Developer Forum, Spring 2002

Pat Gelsinger, Vice President, Chief Technology Officer
San Francisco, Calif., USA
February 28, 2002

PATRICK GELSINGER: Thank you. Good morning. This has been a good week, hasn't it? Anybody who has been to every IDF, all of them, stand up. 100 percent of them. Give them a round of applause. You are my best friends, my very best friends. Even a few of the press folks.

(Laughter.)

PATRICK GELSINGER: No, this has been really quite an event for us. We started with a couple hundred of us. Now we've grown to thousands of us, and it's really a seminal event, not just for Intel but for the entire industry, and from the deepest part of my heart when I walk into Moscone Center, see the big scene, the tremendous showcase, the many of you here I've known for so many years, I want to say thank you very much. It's been a great week.

Today, I get to address you the first time as Intel's CTO. While I've talked to you on many topics in the past in different roles for the company, I'm really excited about today. First time as a CTO. Think about being Intel's CTO. That's sort of what like Gordon Moore and Bob Noyce did.

I want to take this opportunity today, rather than talking about one or two year technologies, to push that window back and take a view for the rest of the decade about some of the things we're working on and crack open the door of Intel Labs and some of our research associates at academia and universities and take a long-term picture of some of the things we're working on that we think might be very significant for us and for our industry looking forward.

To start that out, I have a few of my friends, a few of the luminaries from the industry just to give you their views of Intel and what we're doing, and more importantly, Moore's Law and the role that it plays today and into the future. If we could just play the video right now.

(Video playing.)

GORDON MOORE: I was asked by an electronics magazine to predict the progress of semiconductor components over the next 10 years. The circuits that were available commercially were up to something like 30 components on a chip. And I could see a 60 component one in my laboratory that was going to come on the market pretty soon.

BOB METCALF: The personal computer did not exist. You were lucky to have a computer in a city, let alone a building, let alone a desk.

VINTON CERF: All we knew about were big time shared machines.

GORDON MOORE: I blindly extrapolated that for 10 years, saying we were going to have another ten doublings, we were going to go from something like 60 components on a chip to something like 60,000. That thousand-fold extrapolation turned out to be a lot more precise than I ever could have imagined.

SHANE ROBISON: Moore's Law is probably the most significant underlying principle of change.

JOHN SEELY BROWN: Moore's Law underlies almost everything we do.

VINTON CERF: The silicon technology has played a very significant role in Internet's evolution.

BOB METCALF: As silicon advances advance, there are more things to network.

VINTON CERF: Simply because it's made it possible to build computers ranging from very, very small devices all the way up to enormous super computers.

BOB METCALF: More things to network, which leads to more value.

CRAIG MUNDIE: The advances in silicon technology are the fundamental mechanism that creates opportunity in software. The single most sustained benefit of Intel's work to the Microsoft business is the relentless pursuit of improvement in price performance. It allows us to put the computer everywhere.

JOHN SEELY BROWN: Secure, robust, distributed, low-powered, high bandwidth. The ability to build these new types of fabrics and infrastructures really enables us to build a completely new kind of experience.

CRAIG MUNDIE: We think of a world of many, many devices all with intelligence and software, and all of which are connected together.

JONNEY SHIH: The next era is going to be still a PC centric multi-client era. Both the competition and also the communication and consumer electronics.

BOB METCALF: Every computer communicates and every communication involves a computer. It's the same things.

JOHN SEELY BROWN: When you see our ability to bring communication and computation together, you're going to find fundamental new ways to let the user experience be the dominant form.

CRAIG MUNDIE: The challenge will largely fall onto Microsoft's and Intel's shoulders to provide leadership in this area as well.

VINT CERF: Intel has played a remarkable role in this whole evolution of the technology.

JONNEY SHIH: Intel has prepared to be, again, the best architect for this kind of future.

JOHN SEELY BROWN: Intel has always been at the frontier, pushing the very edge.

BOB METCALF: It needs volume, it needs standards, and it needs to just continue using Moore's Law to drive to new applications.

JOHN SEELY BROWN: Expanding beyond just to focus on microprocessors, to think about the more broad infrastructure that has to be there in order to make many of these new types of capabilities really possible.

SHANE ROBISON: Intel has consistently delivered incredibly innovative technology which is the basis upon which companies like Compaq do the most important things we do.

CRAIG BARRETT: We're just seeing the tip of this iceberg. It's really going to be an explosion, delivering value, delivering new applications and new capability to our end user. That's what our real task is.

(video ends)

PATRICK GELSINGER: Just seeing the tip of the iceberg. Our agenda for today. We wanted to start out by looking at just the extension of Moore's Law, how we're continuing to drive that relentlessly forward. But then to expand Moore's Law, to look at entirely new areas and types of things that we're doing with Moore's Law. Then explore a few specific technologies that we think deliver on this expanding Moore's Law. And then suggest what it might do to computing, to communications, and to the entire geography of our industry.

First, Moore's Law. When Gordon first suggested it, it was a simple observation, like a lot of things. You know, somebody asking to write something down, you stop and think and come up with an observation. After a few years it became a guidepost for our industry and now it's become a law, of sorts, for both technical innovation as well as business innovation, and understanding as a business collectively where technology will allow us to be.

A film director, Francois Truffau once said you have to be a businessman in the morning to be an artist in the afternoon. For us, we deliver technology each morning so that you, our customers, our associates can be businessmen each afternoon.

When Gordon Moore first suggested Moore's Law and we invented the microprocessor shortly thereafter, he went walking around his home and he said where can I imagine semiconductors and processors being used inside my home, and he made 85 observations of where they might pop up.

About five years later he walked around his house again and he observed about 30 more places where they popped up. And over time, we've delivered on many of those technologies and some billion and a half people on our planet now experience technologies that were a direct result of those observations and our collective innovations. But we still have 5 billion people yet to experience the power of the technology and the capabilities that we, as an industry, can and should be delivering to them.

I recall when I was working on the 486 in 1989, we brought out the first 486 chip to the marketplace at a whopping 25 MHz. I was proud of 25 MHz.

I then spent the next three years of my life, all we did over the next three years was going from 25 to 50 MHz with that 486. Three years to get 25 MHz.

Today, in one week we deliver an incremental 25 MHz of performance. In less than five years, in less than one day we'll be adding 25 MHz to our product line. The incredible power of exponential growth.

Gordon once equipped, "If Al Gore invented the Internet, then I invented the exponential."

(Laughter.)

PATRICK GELSINGER: I often get asked where does Moore's Law end? How far can we possibly take this? When I joined Intel in the 1980s we thought boy, one micron, how are we going to get there?

In the '90s we had the 0.1-micron task force, breaking through 0.1. Whew, how are you going to do that?

Today with confidence, we're talking about breaking through 0.01.

I want to promise you, at least until I retire, two and a half decades from now, Moore's Law is alive and well. We are driving the innovations. The future has never been clearer to us than today in advancing Moore's Law. Many decades into the future we are going to be continuing to drive this guidepost, this direction, this law of technology and business forward. There is no end in site.

In this book, Richard Matheson, "The Incredible Shrinking Man," the final line in that movie is "to God there is no zero." And to us as an industry, we have to take the view, the perspective, Moore's Law, there is no end. We're going to keep driving it forward and we as a company and as an industry need to keep driving that core technology forward to accomplish that.

In this last year, we've shown incredible progress against Moore's Law. We've shown some incredible breakthroughs to accomplish that.

Our job is we've continued to melt computers into silicon, and we've done that for several decades now, innovating processors, new functionality coming onto those chips, delivering higher performance, multiprocessing new capabilities for that. We've continued to extend Moore's Law looking forward.

In three particular areas we've shown breakthrough innovations in the last year, and let's just briefly summarize what some of those have been.

The first one has been with our terahertz transistor. At the heart of what we do is building these incredible little switches that are smaller and faster and now we've shown 70 nanometer, 30 nanometer, 20 nanometer, 15 nanometer gates with this entirely new structure, the terahertz transistor structure which is built on three primary technologies, a high gate dielectric, so a new dielectric material moving beyond silicon dioxide for better leaking and switching characteristics.

Also moving past a technology of SOI, we came up with a technology called fully depleted gates which is better, having an entirely depleted substrate, reducing leakage current dramatically, not just modestly, and finally to reduce the resistance and increase the overall performance of the transistor, the ray, source and drain technology.

We think this transistor is a fundamental new architecture for how to build them and will extend us through this decade and into the next with that core transistor technology.

We've also shown entirely new innovations in packaging technology. With our BBUL, our bumpless buildup layer technology. Say that five times fast. Can you do that? It's a bit of a tongue twister. The engineers love it. But here's the bumpless buildup layer technology. Instead of having the die on top, we embed it into the package. You can look at it in the cross-section. And putting it on top rather than into the package with BBUL might not seem like a lot, it has a dramatic improvement in the performance of the package. It's smaller, improves the inductance characteristic of the package significantly. It also lends itself better to multi-chip packaging. Maybe lasers don't make it all the way into silicon and this technology will be better at putting lasers and silicon components directly into the same package. Major new innovation, a technology that we will be bringing forward in the second half of this decade.

We've also done a fundamental breakthrough in the area of lithography, and this is with our extreme ultraviolet program. We realized that the light spectrums we were working on to send them to magnification were no longer scalable. We found as we kept moving down to shorter and shorter wavelength of light that it was just being absorbed in all of the magnification materials we had available. We needed a fundamental breakthrough. About five years ago we launched the industry consortium for EUV and this year we demonstrated the first EUV mirroring techniques. So rather than magnifying, taking through glass, we now use mirrors. So smoke and mirrors. And given the light sources we're using, they're pretty close to smoke. And demonstrating an entirely new approach to lithography. And here is a reflective mask, so instead of sending it through the mask we reflect to absorb the patterns in the EUV technology.

This will go into production in the second half of the decade, and again we're on track to keep driving this forward well past 2010.

So we stand here today more confident than ever for a decade-plus of Moore's Law. Having demonstrated, we've never been in this good of shape. You say what technologies are you going to use at the end of 10 years? We scratch our heads. We know we'll come up with something. We've never been in better shape than today in knowing the future of Moore's Law.

But not only are we working on extending Moore's Law, but we're also understanding the core limitations, the key things that would prevent us from taking advantage of those capabilities as an industry.

At ISSCC last year, we spoke specifically on one of those: Power. We had seen that the early part of Moore's Law was limited essentially by innovation, limited by breakthroughs and moving from the black magic of semiconductors to the science of semiconductors. That was the first 10 years. The next 10 years were dominated by the cost of semiconductors, and moving up the yield and the manufacturing learning cycle.

We predict that the next 10 years are limited by power, and by the limitations that power imposes on the physical form factors that these devices ultimately end up.

So we say Moore's Law is great. We're on track, by 2010 30 gigahertz devices, 10 nanometer or less delivering a terra instruction of performance.

If you look at this curve we've gone smoothly from one watt to 10 watts, 10 to 100 watts, and we're on track to go from 100 to a 1,000 watts. And what's beyond that? 10,000. Exponentials are incredible when they're working for you. When working against you, they're not so nice.

It's even worse than that because power density is you have the square of that, you're trying to put this power into a very small area. And just for fun, we put a few major guide posts, you know, hot plates, nuclear reactors, rocket nozzles, and finally my favorite was the surface of the sun. This wasn't going to work, right? People are not going to carry rocket nozzles on their laptops.

(Laughter.)

PATRICK GELSINGER: Then the last year you've seen us deliver major breakthroughs in architectures to solve this with Hyper-Threading technology for products that are specifically focused at the instruction set level, the implementation, with XScale(tm) microarchitecture to deliver power performance. Also as part of our terahertz transistors announcement, we had major focus on the improved power characteristics of this new structure of transistors as it has dramatically reduced the leakage currents from those devices. And finally, at the recent ISSCC just in January, we showed several breakthroughs in high speed, low-power circuits, body bias techniques that significantly reduce the power of these devices while continuing to improve their performance.

Major breakthroughs.

So extending Moore's Law on track, identifying the limitations of Moore's Law, life is very good.

But we're still left, then, with the question of what else? How else can we apply this incredible landscape of silicon? And what other functions might we apply? What else could we possibly do with it? And that's really the bulk of the topic that I want to take on with our time today.

We call this Moore's Law is expanding. Moore's Law is moving beyond just transistors to become entirely new functionality of devices that includes transistors with entirely new types of devices.

So from transistors to a broad set of devices that include transistors, but also include entirely new structures that we never thought were possible before.

So as we're opening up the lab, you know, cranking open the door, there's a number of technologies. They've got some crazy guys in there, a number of our research associates we're working with, obviously working on new ways to do logic and memory, looking at optical, sensors, wireless, different mechanical structures with silicon, biological structures, different fluidics research, a whole set of research activities are going on when we peek open the door and look inside.

What we wanted to do today is say how, when we open that door and pick a few of these, the power of silicon. When we look at silicon, we now are saying we're going to go from performance to entirely new classes of functionality. When we embed them into silicon we can cost reduce them, move them out of the niches and make them ubiquitous across the planet. We can deliver not just traditional uses but the cost and ubiquity enables entirely new markets to emerge from these technologies. We democratize technology. We make it available everywhere to everybody through the building blocks that we deliver.

So today I just wanted to pick three of these, explore them a little bit more, and suggest their implications on the industry.

In particular, we wanted to look at wireless, at sensors and sensor networks, and finally optical technologies.

The first one, wireless. We have this vision, we'll call it radio free Intel, where literally our efforts are to make radios so cheap, so integrated that you can view them as ubiquitously available. Radio free Intel.

To accomplish that vision we have about five different technologies we're working on that I wanted to touch on briefly today.

The first one we're working on is called MEMS. MEMS is micro-electrical machines. You've probably seen all sorts of crazy things in science and other journals on people building systems using MEMS. We're taking silicon and combining it with this incredible lithographic engine we've been working on to create some cool and innovative things in silicon.

We looked at that and said the area we want to specifically focus on is in RF. We have a lot of things going on in radio frequency work and radios. That's where we want to focus our attention on MEMS and come up with abilities we can productize in the future in the area of radios.

So in our research work what we have done is we have developed cantilevers, so little cantilevers in silicon that we can use for switching, maybe switching antennas. We've also done bridge oscillators for the use of capacitors, filters, selectors, also created inductors for high quality passive components.

What I have here is our first MEMS wafer that we've produced. We've shown the structures that I'm describing here. These are actual MEMS photographs. The one on the lower right for you, you can actually see there's a little air gap. So we're using air, which happens to be a great insulating device to actually create this little inductor and capacitor network that we've built on this silicon wafer that we have here.

We're actively working on these core technologies. In doing that we're going to be able to dramatically reduce the size and cost of the passive components required in radio circuits.

I'm very happy that at the conference today we have the Intel Labs presentations and sessions, and Dr. Bob Rowe, our Intel Fellow who leads our MEMS research is leading those and will go into more detail on the results and technologies that we've accomplished in the area of MEMS.

We also realize that in RF requirements or in wireless environments, there are multiple wireless standards going on in multiple wireless environments. And what we've been working on is a technology in the area of roaming, where people might start out on the wire LAN, might then move to the wireless LAN, might move out of the range of the wireless LAN and want to switch to the wireless WAN. So we have gone on from there. This was at the windows XP launch, demonstrated there, I was told the best demo of the launch getting lots of press. We had it in the showcase here, and we're working on expanding it not just for LAN to wireless LAN, but to make it a complete technology that covers all the way out through wireless WAN that deals with security, VPNs, et cetera, and all of those seamlessly and automatically handled that you're able to keep an application running, all your IP addresses life, when you across that range of connectivity in the future. And all that is underway, in fact, today as we're speaking. So we're driving roaming technologies.

The next area that we're working on is in the area of silicon radios and dynamically reconfigurable radios. What I want you to do now, everybody, I want you to take your cell phone out of your pocket. Now, turn them off like you were already supposed to do. Okay? You've got them all off? Now, the next thing, if you don't have one, reach into the pocket of your neighbor and take theirs.

(Laughter.)

PATRICK GELSINGER: Okay, the next thing I want you to do is to smash it on the chair in front of you, okay? Are you ready? You didn't do it. You guys, when I'm on stage, you're supposed to do what I tell you, remember? That's how it works. But if you rip one of these things open and what you actually see inside is a few silicon components, highly integrated silicon components. But what you see a lot more of is an incredible amount of passive components. You see RF, you see capacitors, you see inductors, varactors, switchers, a whole set of things that limit our abilities to shrink these radios down. And this is an area of research, through MEMS and some of our CMOS radio work, is to focus on integrating the rest of the radio as well.

When you look at the radio today, you've got all these passive components, a few silicon components, radio component, baseband processor, memory component. What we're driving is that we want to integrate those, and actually today it's actually worse than that because if you want to cross multiple networks you need a separate LAN card, a separate PAN card, separate WAN card. We want to drive those in the future, and we think MEMS will be the key technology to do that, allowing us to get rid of all the passive components and put those directly into silicon. But still we need multiple of these cards. We need one for PAN and one for LAN and one for WAN. And the expense, the cost, the power of those is still very high.

So we want to drive all of those together into a single radio component that includes everything, includes the passive, includes the amplifiers, baseband, allowing us to integrate all of those pieces directly together.

This is the area that we are researching and focusing on. What I have here is one of the wafers that just came out of our lab. On this, we've actually built almost the entirety of radio circuitries directly in silicon. CMOS silicon, not SiGe, nothing else like that. But as our CMOS silicon processes have continued moving forward, we've been able to show low noise amplifiers, baseband amplifiers, RF switches, almost all the radio circuits required to deliver a complete radio circuit technology, and these are operating at 10-plus GHz today. Based on that, we're confident that this stuff we can make work and work well at the 5 GHz range for things like wireless LAN and 802.11a.

Furthermore, we're researching the area of software configurable radios, the actual ability to operate on a dual or tri-band, sense the existence of other networks and immediately relocate yourself onto those other networks based on the best network, whether that's driven by cost, performance, or other characteristics.

So these are areas of research, and we're very encouraged by the results that we're getting.

But we don't just want to get to the point of one piece of silicon for an entire radio. We want to drive this to the point where it's nothing but the corner of the die. Where literally, on every chip that we make, we would have a CMOS radio being able to be integrated directly into the corner of the die. Imagine every chipset we built would come with a radio. Imagine every processor that we might build comes with an integrated radio as part of it.

And when I say radio free Intel, this is what I mean. That we literally can get to the point where radios are integrated into every product that we build, able to operate across PAN, across LAN, across WAN environments, seamlessly roaming, connecting to all of them.

We're also working on the next generation the area of ultra wide band technologies. This is spread spectrum technologies, some black magic stuff from way back in the defense days. And instead of having a narrow set of spectrum you're operating in and trying to cram a lot of information, you take a lot of spectrum and put a little bit of information across a very, very wide spectrum.

While technology like Bluetooth* is great as a personal-area network, its bandwidth is severely limited to truly replace wires.

So we've really wanted to drive to a technology that would allow us to truly replace wires and to deliver a fully high performance personal-area network. And we're emboldened that ultra wide band could potentially be that USB-2 replacement, able to operate in the range of ten meters and able to deliver well in excess of 100 megabit per second kind of performance. We've been very active with the FCC and we're very pleased with their recent findings that are allowing purposed emissions across a wide range of spectrum as long as you stay below the power limitations of any one of those spectrums. So we were very active in rallying the industry to participate there.

I'd encourage you, if you're interested in this topic, we just relaunched this week the Intel Technology Journal. It's now an ISDN number as well so it's a fully referenceable journal and there are some seminar papers on the whole topic of UWB as part of Intel Technology Journal. Also another session being led today by Ben Manning is covering some work in the area we're doing in the area of UWB.

So imagine when radios become free, they're fully integrated, every device includes seamless roaming, multi-network connectivity, always connected, able to be across all visions. Imagine where your mobile PC literally is always connected, always able to synchronize, always able to deliver information on any network at any location. That's the vision. That's what we're working on with radio free Intel.

Second area I wanted to talk about was sensor networks, and to help me do that I have David Culler here. But before I invite David up, I wanted to talk to you a little bit about some of the research model that Intel has put in place in the last couple of years.

What we've done is we call it our research without walls model. We've restructured all of our research grants. We have some 500 active research grants that we deliver to universities. We started a whole internal research seed program, and we also put in place what we call lablets. These are small labs, about 50 people at their maximum size, which are led by a senior researcher from the university and who are staffed half and half with university students and personnel and Intel students and personnel in this highly collaborative open research environment.

We've launched three of these now, one at University of California Berkeley, one at Carnegie Mellon University, and one at University of Washington. And I'm very pleased that a person of David's reputation has agreed to come and lead the launch and the initial research activities of our Berkeley laboratory.

David is director of the lab there for us. He was an NSF presidential faculty fellow, and his research area is a focus and interest, specifically look at very small devices, very large networks in a wide range of configurations. And would you please join me in welcoming Dr. David Culler to the stage today.

(Applause.)

PATRICK GELSINGER: Good morning, David.

DAVID CULLER: Good morning, Pat. How are you?

PATRICK GELSINGER: Good. Thank you for joining us here this morning. Just a few thousand of my closest friends.

DAVID CULLER: Just a few.

PATRICK GELSINGER: David, could you tell us specifically about the area of sensor networks and what you're working on there?

DAVID CULLER: Sure, I can do a little of that and I think a few of your friends might have seen a bit of the scalability of this technology in the fall. I thought I'd give you a little of the workings underneath the covers, if you will.

What this allows you to do is spread computing and communication out through things and through space, and it's very energy aware. So most of the time they're sitting there in a very low energy state. If something happens that may be important, you need some information or something happens, they'll first kind of nudge one another and kind of gossip and wake up at a very low energy level. If, in fact, it's important, then what they do is share enough information with their neighbors to build a network out of whatever structure is there. And these are pretty sophisticated algorithms. They pay a lot of attention to how far their neighbors are, the energy and whatnot. And it's not just a matter of devices out there. You can push programs into that so it's a very general-purpose capability. And you can process information inside the network as well.

So usually we have very sophisticated sensors, but in this picture, if you had a green sensor and a red one, you might have the network figure out how many red ones and green ones. Are there more red ones and green ones processing that up inside the network itself.

PATRICK GELSINGER: Thank you for the tutorial, David. And if I were one of your Ph.D. students I'd probably be very interested at this point. But I'm not. Most of these guys aren't. Can we actually do something with this stuff?

DAVID CULLER: It's that Stanford viewpoint. Let's see.

(Laughter.)

DAVID CULLER: So we thought you might want to have some fun with this here.

PATRICK GELSINGER: Okay.

DAVID CULLER: So we put one of these little devices in a bigger box that you can get at with your hands and play with. And you don't mind if we have your friends help us out a little, would you?

PATRICK GELSINGER: They're all yours.

DAVID CULLER: Okay. Well, we did put about 100 of these out there in the audience. Look, they already know where to look. So reach down underneath your seat, usually on the side. They're mostly along the edges. And yeah, you better check the seat next to you. If you grab one, just go ahead and stand up. If you could stand up, hold them high, we can see. There you go.

Now while you were standing up, you can see on your screen the topology of the network that you just built. So you can go home and tell your boss you built a hundred node network in a second and a half. You don't have to tell him you didn't have to do anything.

These are complicated to understand. These really do find whatever information is around. So if we take the spatial framework out of that, you can see what the network actually looks like down inside. So this is how the network sees itself. You can see that it's constantly working and adapting to small changes in the radio environment.

PATRICK GELSINGER: So it's actually dynamically reconfiguring itself.

DAVID CULLER: That's right. It's pretty stable right now which says we have a nice quiet radio environment in here.

PATRICK GELSINGER: And none of our people are running around a whole lot.

DAVID CULLER: That's right.

PATRICK GELSINGER: Sort of standing still. Can we liven things up a little bit?

DAVID CULLER: Maybe we should make something move?

PATRICK GELSINGER: Okay. Let's do that.

DAVID CULLER: I brought some other nodes I could throw out into that so you could get a feeling for that.

Why don't you toss some of these around. It takes a little while for those nodes inside the balls to find where they're next to and start routing information back to us. So it may take a moment. They're moving so fast that the network is having trouble seeing them. I guess if any of them actually stayed put.

PATRICK GELSINGER: These guys were at the party last night.

DAVID CULLER: Think that's right. I think they're having more fun with the balls than with the network anyway.

PATRICK GELSINGER: Okay, okay. That network is really nice David.

DAVID CULLER: There they go.

PATRICK GELSINGER: Now you see it pop up, getting further back in the room. So you see the dynamic, self-configuring, reconfiguring network, but could we do a little project with this with our audience?

DAVID CULLER: You know, we can do lots of processing. Is there something you'd like them to do?

PATRICK GELSINGER: Since we have red-green, we could take a vote, right?

DAVID CULLER: You could get a sense of the network. It's a primitive sensor, button sensor.

PATRICK GELSINGER: If we can bring up our question, we have a little experiment and I want your feedback in this. We're going to pick the location -- now, the '02 IDF locations are already set, but '03, I would like your sincere feedback in the location for the spring IDF in '03. And if you press the red button you're selecting Hawaii. More people are getting into Hawaii. And if you press the green button, that will be Silicon Valley. Okay? So we're either back in Silicon Valley somewhere or Hawaii.

Are your bosses going to approve Hawaii? Not a fat chance. Say it's an Intel Developer Forum in Hawaii? No problem.

You can keep voting. This is like proxy votes. You can vote early, vote often.

(Laughter.)

PATRICK GELSINGER: Okay. Now, where do you think the vote for when the next IDF is going to be? Yeah, okay. Silicon Valley. There we go. Some of you really think we're going to go to Hawaii? Fat chance.

Okay. Thank you very much, David. I appreciate...we have one more thing. I almost got you out of here too fast. Keep going here, back to the slides. Silicon Valley, 70 percent. That's really where we will be.

So that's a lot of fun, we enjoy it, but what do you think -- while our audience enjoyed this, it's not really a real-world application. How do you think people will use this, David?

DAVID CULLER: There's a lot of exciting applications for it, really. Many of them go outside of traditional environments. Safety, a very important one. You can imagine baby cribs that could listen to your child's breathing. Pools that tell you when things fall in. Tremendous efficiency in manufacturing. Imagine if you could put a weather station in every plant or in your manufacturing environment that the motors themselves or the structures tell you when they're ailing. So very interesting applications in many areas.

PATRICK GELSINGER: What do you think are some of the research challenges that we need to deal with and attack in this area?

DAVID CULLER: Some of what you touched on in the hardware side in really bringing this down to low-power structures.

PATRICK GELSINGER: These things were pretty big yet, weren't they?

DAVID CULLER: Well, inside that box they're something about this size, and today we can easily squeeze it down into something the size of a quarter. But going forward, this will vanish into a spec of silicon.

PATRICK GELSINGER: Perfect, perfect. And that's our job, right? That's why we're here. We're going to take these radio free Intels and smash them into a spec of silicon and put nodes everywhere in the planet. Thank you very much, David.

DAVID CULLER: Thank you. Thanks for your help.

(Applause.)

PATRICK GELSINGER: So we've looked at wireless, the wireless area, radio free Intel. We've looked at an application of that technology, and also of an entirely new class of sensor networks and some of the MEMS applications that we'll touch on in the MEMS class today will touch on some other sensor applications where we've crossed the atoms-to-bits boundaries.

The third area I wanted to look at today is photonics or optical networks. I've been going through a little bit of an education on what this optical stuff is all about.

Now, for the few optical experts in the audience, I apologize, but let's go through a brief optical tutorial, and then we'll look at a specific technology or two we're working on here.

So optical, you start with some light source, typically a laser of some source that you pick a specific wavelength of light.

And then you modulate that. You put information on it, and on an individual wavelength of light we might be able to put 50 Giga bits or more modulated onto a single wavelength of light. And the beauty of fiber net networks is you can go almost lossless for an extraordinary distance, tens or hundreds of kilometers. And at the far end of that network you go into a photo receiver which has some filter which picks out the specific information you're interested in, and then a diode, which demodulates it and pulls the specific information off of it that you modulated onto the first place.

So that's your basics of optics. Sounds simple.

Well, then the next thing that happens is wave division multiplexing. The beauty of an optical fiber is you can put lots of wavelengths of light onto a single piece of fiber and they don't interfere with each other. And wave division multiplexing is you multiplex many of those together, may be ten, 50, people are researching hundreds of wavelengths of light onto a single fiber. You go into a multiplexer, shove it across this fiber that may be across the ocean or nation on the other end, demultiplex it, choose out the particular light that you want, demodulate it, and you're off and running. With wave division multiplexing, we could see literally on a single piece of fiber pedabits of information being transferred.

Research is then saying today those are fixed frequencies so we're actually able to make them tunable. So you can have multiple common hardware for both the laser and modulator as well as for the demodulator that will drive down the cost and allow further reduction in the operational costs of optical networks.

What we want to look at specifically is driving the radical reduction of costs through silicon integration and the use of greater manufacturing technologies to deliver drat tick orders of magnitude reductions in the cost and the ubiquity that results from that by applying Intel photonic silicon technologies.

Let's take a look at those. This week you saw Sean Maloney talk about our new products in the area of optical interconnects, and what we've done is taken our high speed CMOS and silicon germanium work from other process technologies and drive the integration of the electronics portion, but also with some breakthrough innovations in manufacturing technologies, dramatically reduce the size and thereby the cost of these core components. What we have here is called a pill in the industry, this small little butterfly and driver. We've integrated those together. Today these technologies might cost a couple of thousand dollars. This piece that we're looking at in front of us is a couple of hundred dollars. Driven an order of magnitude reduction in this particular portion of the opticals products.

Using that technology, we've been able to continue to drive the final product reduction and cost. Here I have a transponder. This is the product that we've been shipping for a while. This product in the past would have been $40,000. This particular version was about $10,000. The product we introduced this week, less than $2,000. And in the future, we want to drive these transponders to be well under a thousand dollars.

We also introduced a 10 gigabit Ethernet that uses a 70 pin XAU interface, 10 gig Ether receive and transmit technologies. In the past this would have been over $10,000. What we've introduced now is $2,000 and our goal is to drive these to less than $500 into the future.

Now, these are the products that we have today, and you see orders of magnitude reduction in the cost through integration of silicon and improvements in manufacturing technologies in order to get that far. But let's open the lab again and go a whole lot further.

What we have here is an optical generator, spectrum generator, and we're generating a set of frequencies which are multiplexed, and we're feeding into our little test setup here, so we have a number of wavelengths of light that are being fed into our piece of silicon here. And this silicon, we have a Bragg grating filter that we've manufactured in silicon.

The electronics you see here are to enable this to be tunable. So the actual filter here is simply being tuned, and you can ignore the rest of that. This is where the action is up here.

Out of this tunable filter, then, we have the filtered light that we're now looking at. And coming over here, this goes into an optical spectrum analyzer, and from this we're able to look at the filtered results and look at the resultant waveforms that come out of our optical spectrum analyzer.

On our screen here we simply have a Windows face of exactly what's being shown on the optical spectrum analyzer and a little GUI that allows us to tune the filter.

So I'll come over here and right now today we're sitting at 1551 wavelength, 0.72. Let's go to 1552.52.

Today this is typically a mechanical process where you actually have to send somebody out and reset the mechanical optical mechanical switch to do so. So you're talking about a long time to reconfigure a network. But as you saw, we reconfigured our network in just a few seconds. Let's refilter it. We'll go up here to yellow or 1553.32. Just a few seconds later, we've retuned our filter.

Pop it up here to red. Back here to blue. You can see we're dynamically reconfiguring our filtered spectrum of light. If you're close enough, you can see this is actually what's going on.

Now, the power of this, it's really incredible because this is the mechanical product, the mechanical filter that is shipping today, $10,000 device today.

What we've done is we've built the core of that, this filter, directly into silicon. This is using polysilicon technology. We've created a Bragg grating filter out of that and using some defraction properties of silicon created a filter in this little test chip that we have here. Better yet, it's not just one filter that we put onto this test chip but it's literally a hundred. And this came out of our Intel fab. I'll show you the test wafer that we're actually showing here. So on here we have about 100 filters. And how much does it cost for a little chip like this? It's pennies per filter versus $10,000 in an example wafer that we produce from our fab in this particular area.

So overall by applying silicon technologies we're driving down and increasing the integration of silicon optical capabilities. The results are ten, hundred-fold reductions that are going to result from these innovations in optical technologies looking forward.

Our goal is to combine the power, the strength, the innovations of our silicon electronics technologies, putting those together with our research in these core building blocks, wave guides, filters, what we just demonstrated, modulators, fast switches, innovate those core building blocks and then pull those together into an integrated electronics and photonics capability. The results are dramatic reductions in the cost, the ubiquity of optical networks and optical connections in the future.

We've briefly gone through three specific technologies. We've looked at wireless, looked at sensor networks, and looked at silicon photonics. Each one of these we've covered on briefly. There will be a little bit more detail on a couple of them in the lab section today. Over the next couple months we'll be releasing white papers in more detail on these specific areas of research and I encourage you if you're not part of the Intel developer update, we have the URL on the screen here and you can get updates on these areas of research we have underway today.

We think that the implications of these technologies are very significant. Over the course of the week you've seen a lot of -- you've seen Intel talk about a lot of different things we're working on. A lot of the work we do at Intel Labs is trying to bring these areas of our business groups and architectures together, trying to drive common architectural elements across IXA and across IA-32 and Itanium. You've heard 3GIO and its application across that whole set. We've talked about back plane technologies. We've talked about how optics and our innovations in our communication product line are going to roll in and be integrated directly into our chipsets and processors.

We think that silicon photonics is going to be a breakthrough technology that will allow us to bind them more together.

Let's be Gordon Moore and look at the implications of silicon photonics. Today we have optical networks almost exclusively in the local-area network across the nation. We see those coming into the enterprise, all the way into the data center, the backbone of the data center, the connectivity between racks, between boxes, and eventually all the way to the processor itself.

We can imagine, as Larry Smar does in a research project from U.C. San Diego, the opti-puter. We're working on the opti-processor where we're bringing it all the way to the processor itself and you can imagine the disaggregation of servers, the role of networks that will happen as we bring those entirely together, as we bridge from photons to electrons and bring the power of Moore's Law to optical networks.

You've also heard us this week talk extensively about our PC activities and also our handheld activities, and you've seen many technologies that were common across those two. We saw our compiler technologies were common. We've talked about many of our libraries and developer tools. Security and roaming technologies, again, common across both of those. We think again, wireless networks will be, and silicon radios will be a bridge to bring those even closer together into the future.

Again, walk around with me and imagine what will happen when we have radio free Intel, when we're able to fully integrate wireless communication into each of our devices.

We've talked for a while now about this notion of an extended PC where the PC delivers services to other devices in the home. It will be realized when we deliver integrated radios into every CE device of the home and PCs are able to connect and deliver services to those devices everywhere in the home.

Imagine the cell phone of the future is no bigger than an earring. We were discussing some things and Martin Reynolds suggested the microphone button where, with a wireless network, a MEMS microphone device will be no larger than the size of a button on your shirt. You'll be able to have a microphone rather than this big thing hanging on my belt. We'll actually deliver a wireless world where you no longer need wires to connect these together. We want to bridge from the RF world to the electronics world and bring those together with the power of Moore's Law.

Let's walk around again and imagine the world that's possible with ubiquitous, very inexpensive motes that are being delivered that are being integrated into every room and device of the home. You can imagine products that are able to identify and maintain themselves, notify when different devices need to be upgraded or looked at. You can imagine smart farmlands where literally every vine plant will have its own sensor specifically making sure that it gets exactly the right nutrients, exactly the right watering. Imagine the impact it could have on difficult areas of the world for agricultural purposes.

Had a friend recently who is rather elderly, and her knee -- when you lose all the ligaments in your knee you have excruciating pain. She was walking into of the bathroom, turned the wrong way, excruciating pain and fell into the bathtub and was there home alone for hours. Imagine a world of smart detectors who could have detected and notified that. Imagine the smart infant blanket that can detect heart rate, et cetera.

As we bridge the world of atoms to bits, imagine the implications it will have on us, our industry, and the world.

As we look to these areas of innovation that we've discussed this morning, Intel's obligation is to continue to be the building-block supplier. Our industry is built on the continuing innovations of Moore's Law. That's what we do. We deliver building blocks, then, that enable your innovation.

Our job as an industry is to continue to work together, to take advantage of these innovations and these breakthroughs to allow us to develop new technologies, new products that change our world.

Finally, my commitment is to continue to lead Intel to be that open, inclusive industry colleague that you could have relied in the past and that you can partner with in the future that will result in creating opportunities for all of us.

Gordon Moore said that Moore's Law was changing. In many ways, it's still the same. It is still the guidepost. It is still the metric that we judge our industry by and our continuing innovations for years and decades into the future. But in many regards, it will never be the same again. It will be so much greater, so much broader, and impact so much more of the world than we've ever envisioned as a result of these technologies.

We see the world of an expanded Moore's Law as bigger, better, more profound opportunities than we've ever had before. We come from a finite past where we had finite resources. We're looking forward to an infinite future where we have infinite resources that we collectively will use in infinite ways. That's the promise of the expanded Moore's Law.

Thank you very much.

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.