Intel Developer Forum, Fall 2002
San Jose, Calif., USA
September 9, 2002
ANNOUNCER: Ladies and gentlemen, please welcome President and Chief Operating Officer, Intel Corporation, Paul Otellini.
PAUL OTELLINI: Thank you, and good afternoon. On behalf of Intel, welcome to our Fall Intel Developer Forum.
If you watched that video, you could look at those quotes and wonder if they were isolated instances or if perhaps they were footprints in the snow in terms of where the industry is going. I tend to think them as the latter.
The industry is changing. The technologies we've been working on for years are coming together. The investments that Intel is making and others in the industry who are investing through this toughest retrenchment in the history of the industry will pull us through ultimately with new and exciting products.
These products will deliver, I believe, the fuel for growth for reigniting the industry.
Where will this growth come from? I think very simply you could think about the fact that we are about to usher in the third era of computing.
First era, some of us grew up with. Those of us with some gray hair remember the mainframes. Mainframes were all about batch computing, glass houses, computing isolated from the users. But it worked.
The second era was built around the personal computer. It brought forth interactive computing in a networked environment.
So is there a third era? And if there is, what does it look like?
I think it's very likely to be built around a series of converged devices that integrate the capabilities of computing and the wireless communications infrastructure that the world has been building together into devices that give you the best of both.
And collectively as an industry, we are poised to deliver these devices, and that's what this speech, and that's what this conference over the next three days, is all about.
The ultimate goal, the thing that we at Intel are working for -- and let me take a stretch and say the things I think that all of us in the industry should be working for -- is really bringing computing to everyone anytime, anyplace in the world.
Now, the question you might ask about that is why now? It's been talked about for some time before.
I think it's time because computing is increasingly more pervasive. It's about Moore's Law delivering the cheaper local intelligence that's increasingly available and delivered in the microprocessor as it becomes ubiquitous.
And it's not just in the PC. The same intelligence is being built into servers in various configurations, into handheld devices, into cell phones, into network processors, and changing the telecommunications infrastructure by bringing the benefits of standard computing platforms to that industry.
The second question you might ask is why Intel. Why should we be one of the companies to make this happen? I think very simply that what this is what we have always been all about.
Intel is a 34-year old company this year. Our heritage is deep with integrated technologies. It's what we do. And in fact, the name Intel is an amalgamation of the two words -- integrated electronics. It's what we have been focused on from the beginning. It's what we're good at. And we believe it's what will drive this next era of computing.
Many of you know us as the microprocessor company, but we're more than that. We moved to segmented models for our processor and chipset development some seven or eight years ago, and those segmented models allowed us to move computing and tailor it increasingly for the needs of users. Today, we do that in terms of form factors, in terms of power, focusing increasingly on usage models and also on scale, scaling out and scaling up, to provide the server infrastructure for enterprise computing.
We are also the world's leader in communications silicon in terms of revenue, in terms of volume, leading in Ethernet, leading in handhelds, leading in silicon that goes into wireless phones today.
So if you look at where we have been, what we are good at, and what we have today, and if you believe that converged devices will usher in the third era, then you must conclude that silicon is the engine for that happening.
In many ways our history reflects this. One of the early successful products at Intel was the DRAM, which essentially replaced core memory in mainframes with solid state memory.
The microprocessor, which was invented to replace a number of discrete components that did mathematical functions in a calculator, became the heart after personal computer, and ultimately its architecture changed the mainframe and server environment as well.
For this next era, we will take our architectures, we'll take our silicon capabilities, the things that we are building into 90 nanometer silicon, technologies like strained silicon to improve the performance of a given transistor, add those all together and create integrated devices for this converged world.
In doing this, we are going to exploit a very simple proposition. You saw it on the video. All computers will communicate, and all communications devices will compute. Therein lies our opportunity, and as developers in the same environment, your opportunity as well.
Our mission in this -- to bring forth the notion of convergence. Pat Gelsinger, Intel CTO, mentioned it as the overarching theme. This is something that's been growing in the Intel commentary over the last couple of years. As we've been able to start thinking about how we'll integrate the two basic thrusts of the company, communications and convergence. And I'll talk more about that.
But in this concept of convergence, we have a mission. Our mission is to drive or to accelerate that convergence simply through silicon-based integration. It's what we do well.
Now, you can go back and say that people have been talking about this concept of convergence for 20 years. It's been the Holy Grail of computing. In the beginning, and indeed they were far apart, there was the communications industry, built around primarily analog technologies, large switching centers, sometimes taking up an entire square block of a city, half of it full of batteries for backup in case the power went out.
There was the computing industry that grew up around mainframes and glass houses.
But as Ethernet and the early local area networks were deployed, the ability to start thinking about a vision of combining these two industries, using one to benefit the other, started taking hold.
In fact, the notion of convergence on the quote goes back over 20 years. That's a quote from a Canadian newspaper talking about ushering in the age of convergence circa 1979.
If you fast forward from 1979 to today, you can see that we've made great progress along this dream. The Internet, moving increasingly to broadband capabilities, increasingly to highly visual digital content over those networks, is starting to take place. Each of us increasingly uses in our daily lives devices or applications that are absolute manifestations of convergence. They couldn't exist without it. PDAs, cell phones, Napster before its business model broke, Blackberry, PVRs, telco appliances, and network processors are standardizing the telco infrastructure.
These are all happening today, used by us today, used by, most importantly, people outside of the industry today, moving beyond the early adopters into the mainstream.
So you can start thinking about what's tomorrow. What do we have to do to be able to deliver the device that finishes this convergence, that gets us to this anytime, anywhere device that brings the needs that we're looking for?
Well, it's going to involve new devices, home servers, ubiquitous wireless, modular building blocks for telecommunications infrastructure, allowing you to have devices that allow storage anywhere. In making this happen, Intel has two roles. The first is to deliver the converged silicon and platforms to make it happen, the things we've historically been good at.
The second is to provide the development community with convergence tools, tools that improve the development time, the development environment around Intel's silicon, but that work within today's existing development environments.
In a very simple sense, the goal we have is to allow developers to develop once on an Intel device, principally a PC or a workstation, and be able to have that application then run best on a variety of Intel architectures, optimized for a variety of different kinds of devices. A new value proposition for developers.
This wasn't needed in the early days of the industry. In the early days, people targeted applications around a specific platform. There were no cross platform requirements. With the development of client/server architectures in the '80s, we began to see the opportunity to create IDEs or integrated development environments. Some worked; some didn't. Some still have juries out. OSF didn't make it. Win32 APIs were wildly successful. Things like POSIX we don't know.
But the value proposition has changed. The need to have cross device development platforms, cross architecture development platforms is, I think, one of the limiting factors, because at the end of the day, we're not going to be limited by silicon. We're going to be limited by the ability to create an environment and applications that take advantage of the wirelessly connected silicon that we are about to deliver.
So our value proposition to you as developers in this extended era is really to give you the common denominator of Intel silicon as a target platform to do your development around. That sounds a bit like science fiction. In fact, it's not. It's real today. I have an example here from Macromedia, who does Flash, as you know. In this example, Macromedia has three separate applications where they were developed on a PC and for a PC, but the applications using are Intel IPPs or Integrated Performance Primitives, which you'll hear a lot more about tomorrow from Ron Smith. Those applications could be ported very quickly and seamlessly from the PC to the PDA environment. In this case, they were XScale-based PDAs. And Macromedia has a messaging application that does that now. They have, on the right hand side of the slide is a online banking application, and on the bottom of the slide is an online e-Commerce application, real applications they've done for real companies, allowing people to take advantage of this proposition of developing once on an Intel PC and using it across into multiple architectures and environments.
So I think that from a developer's standpoint, we have a simple role in this as well. The role is to drive convergence. The tools that we deliver to the software community today are up on the slide. They're the libraries, the IPP libraries I spoke about a minute ago, compilers, Vtune(tm) analyzers, tuned for our various architectures, PCA, IA32 and Itanium(r) architecture.
In 2003, we're going to change those offerings and expand them. We're going to create integrated SDKs or software development kits, that allow for seamless cross architecture development compilation, optimization, and deployment, allowing you to move from IA32 to PCA or from IA32 to IXA or vice versa. The same kind of tools we've already brought to the market when moving from IA_32 to Itanium or vice versa. To augment this, we are announcing this IDF the beginning of a software college. This college will take place on a year round basis. Intel software experts will help train you, the development community, on these specific tools in how they work in your existing development environments. We are not trying to change the environment; we're trying to make it better and easier for you along the way.
Now, I think your role as software developers and application developers in this, and to some extent even hardware developers, is twofold. You need to increasingly focus on developing device-aware applications, applications that know what device they're being run on. Am I on a PC for a rich environment? Am I on a PDA with a limited screen capability, am I on a phone with an even more limited capability? How do you scale those applications? How do you repurpose the data types for the device, the specific device that's going to be viewing them workout writing multiple levels of applications.
That's number one.
Number two, I think that you need to focus on increasingly creating applications that are mobile aware. It's not like going out and creating applications for a mobile device, all applications are going to be mobile. All applications need to understand whether they are connected full time, occasionally, and be able to handle that. That's the kinds of applications we think are necessary in 2003 and beyond.
Those of us who live in the wireless world today know that you often talk as much about being connected as you do about being disconnected. The kinds of error messages on the screen are common to all of us road warriors. We have made good progress as an industry here. There are a number of mobile aware solutions that are out there. And I'll point to Excel* or Windows XP* as very good examples of that, where they understand whether you are connected or not and the fact that you will not always be connected, and adjust for it.
As we develop into the world of XML, there's an opportunity and a threat here that we need to embrace. XML is the reason that will give us a Web services architecture that is a prerequisite for this fully connected environment. And we need to be able to develop these apps to be able to be smart around this, develop for mobility.
If we do this, we will be able to start thinking about the next level. At the last IDF, Pat Gelsinger talked about Radio Free Intel as a notion that was out there, something we could do with the transistor budget. We've taken our thinking and gone a little bit further. We're now thinking about the opportunity that a one billion transistor template allows, what kind of chip can you make with a billion transistors. Towards the end of this speech today, you'll see a demonstration of a new server microprocessor called codenamed Madison that is approximately a half a billion transistors. We are in development at Intel today of a billion transistor microprocessor. This stuff is not rocket science; it's not that far away.
When you move beyond the realm of servers, though, this billion transistor template becomes very interesting. You can start integrating logic, communications, graphics, memory subsystems, security algorithms, giving you the capability to do more and more on a single device. As we do that, we start addressing some of the needs that users have told us they need us to work on, self-detecting, auto-configuring, connectivity for anyone, anytime, and anywhere.
I spent some time talking about the vision. I'd like to shift now and talk about where we are today and show you the state-of-the-art of the products today. To do that, let me bring out Bill, who's going to take us on a tour of Banias and Starbucks.
(DEMO begins and ends.)
PAUL OTELLINI: What we tried to show you in the last set of demos is how these things come together. What's the usage model associated with the kinds of products that we're talking about.
We need products, we need applications, we need services, and we need quality of services, which was Jim's last point. That's the use model. There's a lot of hard work involved in making that happen. And what I'd like to finish up today -- this is the last part of the speech -- is talk about the kinds of things that we're doing as a company at Intel to drive this platform.
And I'm going to use a cube as a construct to talk about it.
I'll talk about our foundation technologies, I'll talk about the use model thrusts, and I'll talk about the core competencies of the company in making all this come together.
I want to start with Moore's Law and More Performance.
We said that gigahertz is an important thing but it's not the only thing. On the other hand, gigahertz is a very critical element in continuing to drive performance over time. It's been true for 30 years and it's going to continue to be true.
What I want to do is show you gigahertz is far from over and maybe take you on a tour backstage to a very secret room and show you what Art has lined up to show us in terms of advancing Moore's Law.
(DEMO begins and ends.)
PAUL OTELLINI: Almost 2 GHz faster than our fastest production product today on 0.13-micron silicon. We're going to continue to advance Moore's Law, we're going to continue to drive gigahertz. But in addition to that, we will add performance as we talked about in the last IDF's speech, by bringing you performance on a different vector, using architectural techniques to bring usage-based performance to the marketplace.
One of those trends we talked about last IDF was Hyper-Threading Technology. Hyper-Threading Technology is a very new technology for Intel. The first time we ever demonstrated it was one year ago this week at fall IDF.
We are in production on HT now, Hyper-Threading four server product. We went into production the first quarter of this year, it provides almost 30 percent boost to applications in a server environment.
Today, we are announcing that we will ship Hyper-Threading Technology on the desktop coincident with our 3.0 GHz language for Pentium(r) 4 processors in the fourth quarter of this year. That is an acceleration from where we were in the past. A lot of that is the stability of the product. It also reflects the stability of the software environment into which we're launching this.
Hyper-Threading Technology can provide up to 25 percent performance in existing application and threaded environments. And it's a ground breaking technology.
What we talked about when we described this for the first time a year ago was it allowed us to do more with less, essentially, free performance, two processes inside of one.
To give you an example of how this works on the desktop, let me ask Elise and Diana to take you through the nuances of Hyper-Threading Technology.
(DEMO begins and ends.)
PAUL OTELLINI: Like any new technology, there's a call to action to the industry as well with this one. In order to make threading happen on the desktop, we each have our respective roles. Intel has to ship a lot of silicon. In fact, what we're saying today is we are committed to shipping about 25 percent of our performance desktops with Hyper-Threading next year, on top of about 60 percent of our workstation chips and about 80 percent of our server chips. So we are producing very large amounts of silicon that is Hyper-Threaded enabled for you.
From a developer standpoint, this is a huge opportunity. In the past, had you to wait for new technologies to come into the market very slowly. And that led to an environment where applications were traditionally written to the installed base, to the lowest common denominator. With the scale that Intel is operating on today, the ability for us to switch production to new technologies very quickly, we can start moving new technologies into the market in almost a square wave fashion. So this 25 percent of performance desktops next year may sound like a small number, but then you think about that in terms of the reality of the number, and it's over 20 million desktops, a huge opportunity for developers out there to take advantage of this free performance.
You need to think about this in terms of building your applications, taking advantage of the tools, the compilers, the thread checkers, VTune which is a tool to analyze performance in this environment, and tuning those applications to take advantage of Hyper-Threading Technology.
This technology will be in all Intel microprocessors over time.
Today I want to introduce a new technology, though, just like we did last year with Hyper-Threading Technology.
The new technology has to do with safer computing. This technology is codenamed LaGrande Technology. LaGrande is all about creating a safer computing environment.
If we are going to enable convergence, if we are going to enable the promise of e-Business, we have to be able to have a more secure environment here. Hardware-based strengthening to this is critical.
LaGrande delivers a hardware-based foundation for security. It includes protected execution, protected memory, and protected storage.
It will be delivered into the marketplace through our processors and our chipsets working together to create a secure hardware environment.
We are working with the industry to do this, not just our OEM partners and ISV and OSV partners, but also the privacy experts to ensure we do it in a way that is acceptable to the norms of privacy nowadays.
It's a core technology that things like the Microsoft Palladium initiative can take advantage of to build much more stable platforms.
To give you a more detailed explanation of LaGrande Technology, I'd like to run a clip by Richard Wirt, an Intel Fellow, talking about it.
(Video plays and ends.)
PAUL OTELLINI: This is an exciting new technology that you'll hear more about over the course of the week and in various keynotes and classes.
There are other elements of the platform that I'm not going to discuss today but you'll also see in further keynotes and classes that have to do with platform enhancements. In graphics, power optimization, wireless connectivity and so forth they really are the subject of much of the next few days.
I'd like to shift now and talk about how these technologies are being used. The first place that we've been focusing quite a bit of effort on the last few years is the enterprise, specifically high end enterprise computing, what's referred to affectionately in the industry as heavy metal.
I'd like to bring out Mike to talk to you about the advances in Intel's heavy metal.
(DEMO begins and ends.)
PAUL OTELLINI: I'd like to add one fact to the NEC product on the end. Even though, as Mike said, that was now populated with Madison microprocessors, NEC has a version that has Itanium 2 processors in it. And today NEC announced a TPCC performance of 308,000 transactions on their system. This is the world's fastest SQL system. It doubled results versus any previous SQL results. It establishes a new level of price performance, bringing Intel Architecture into the very high end of TPC results. It's the fifth fastest non-clustered transaction machine in the world itself.
I'd like to wrap up and talk a little bit about the construct you're going to see over the next few days. We showed you a little bit about the enterprise in terms of products, but we only talked about servers.
Intel has changed its product planning process over the last couple of years to start moving much more aggressively towards market-based, market-driven planning. In the context of doing that, we started rethinking how our products are being used.
In the past, you would have seen in different IDFs from Intel with form factors for telecom equipment and so forth. We're changing that starting with this developer forum. And it reflects the way we've done our product planning. Instead, the keynotes you're going to see and many of the classes you're going to see are going to take a different slice of the world, a slice that is much more pragmatic in terms of how people use computers. In the enterprise, you're not going to be looking at enterprise computing but digitizing the entire office along with it.
For the home, the impact of the digital home and how the computer will interact with other digital devices in the home.
In mobile computing, we're not just going to talk about notebooks, we're going to talk about mobile Internet clients, how they work together, what the development environment is, and what the value proposition is to develop on a common architecture.
We'll talk about telco infrastructure and how that is moving much more aggressively towards standards as well.
This is the construct that you'll see in the keynotes over the next couple of days.
Lastly, I'd like to point out that there are corporate capabilities, corporate capabilities that you take for granted about Intel in terms of validation, in terms about safety and quality and reliability, manufacturing scale, technology capabilities. This slice of the cube is really going to be discussed on Thursday's keynotes by Pat Gelsinger and Sunlin Chou as they dig deep on the current and forthcoming Intel technologies.
The combination of all these really gives us the collective opportunity I talked about at the beginning of this speech, taking advantage of all of the development activities that you're doing, that we're doing, and the billions of dollars that we are collectively putting into these new products and maybe creating the opportunity to drive convergence happening a little bit faster.
One way I look at this cube is, it's an engine. It's a convergence engine that's allowing us to produce the products that hopefully together we usher in the third era of computing.
Thank you very much, and enjoy the conference.
* Other names and brands may be claimed as the property of others.