Comdex Fall 1996
Andrew S. Grove
November 18, 1996
[This summary was prepared from Dr. Grove's speaker notes. It is representative of the themes and content of his speech.]
This is a very special year for us as we celebrate the 25th anniversary of the invention of the microprocessor. It is a pleasure to address our industry on such an occasion. We as an industry have made tremendous progress in a short 25 years. The rapid progress that the microprocessor and the PC revolution have achieved may be unmatched by any other technology or invention in history. Today I will illustrate the results of the microprocessor revolution thus far and take a look at where this "revolution in progress"
will take us in the next decade or so.
History of the Microprocessor
In 1971 Intel was a memory chip company. We had a customer, Busicom, who wanted us to help them design and build a calculator. As most of the better known calculator companies already had contracts with other suppliers, we were happy to get the chance to work on this project. Four Intel engineers -- Les Vadasz, Ted Hoff, Stan Mazor and Federico Faggin -- were given the charter to work with Busicom. Together they packed 2300 transistors on a single piece of silicon and made technological history. It was the world's first microprocessor -- the 4004.
[The rest of this section is a videotape documenting the history of the microprocessor from 1971 through 1996. This video is available from Intel Press Relations.]
Today's Connected PC
What started as two very simple applications -- a spreadsheet and a word processor -- has evolved into today's connected communications applications. Today's "connected PC," based on a Pentium® processor and
communications enabled applications, allows users to reach out through their PCs across boundaries of space and time.
Here's an example of how connected PCs are being used today:
Demonstration: Starbucks Coffee Company
This demonstration shows the connected PC in both business and consumer use. In the opening segment, the connected PC is used to communicate across a prototypical corporate intranet from Starbucks Coffee headquarters in Seattle to its remote store locations. Employee training takes place via live multicast conferences led by Starbucks CEO, Howard Shultz. Then Dr. Grove is joined by John Williams, Brand Development Director for Starbucks. Together they illustrate consumer applications through the use of internet technologies to shop for Starbucks products, receive email, and make a videoconferencing connection using Intel ProShare® technology to other Starbucks stores. The same infrastructure that keeps the company's geographically dispersed workforce in touch with their CEO and their culture can at the same time offer the connected PC as a service to their customers.
Starbucks is just one example of how much the PC has impacted our daily lives. Underlying all of our experiences is still a very small device -- although now containing millions of transistors -- a very small sliver of silicon known to all as the microprocessor.
In 1989, we commissioned a team of Intel's best microprocessor minds to publish a paper in the IEEE Spectrum based on our experience building the 486DX microprocessor. The paper was titled "Microprocessor 2000." At that time we projected that the microprocessor of 1996 would look like
Transistors 8 Million Die Size 0.8" (1800 mills)
Line Width .35 micron
Here is where we actually are in 1996:
Transistors 5.5 Million
Die Size 0.6"
Line Width .35 micron
We reached -- or surpassed -- all we set out to do with one important exception: we did it all with fewer transistors. (This is significant because, as the number of transistors used in reaching a specific performance goal gets lower, so does the cost of the device.) Intel's success was due in large part to our focus on improving the microachitecture of the processor.
Evolutionary Performance Improvement
And architecture improvements are still a significant focus of our efforts. The most important element of performance improvement has been the ability to get more data through the processor per clock cycle by exploiting the techniques of parallelism. In the Intel486™ generation, we recognized that performance improvement was not limited by the silicon, but rather by the single instruction path. In the Pentium® processor generation, we increased the instruction parallelism by adding what is referred to as a second "pipeline" -- a second lane for instructions to pass through the processor -- thus speeding up the performance and delivery of data. However, this had a drawback: the instructions in each pipeline needed to be synchronized and completed in the correct order or they would stall before arriving in memory. In the Pentium® Pro Processor generation, we extended the paths to three instructions in parallel and we solved the synchronization problem by inventing Dynamic Execution. Dynamic Execution added the capability for instructions to execute at their own pace and even out of order. The instructions, once executed, all wait in a buffer area from which they are written to memory in the correct order. It is the continual evolution of techniques like these that will further the performance of future microprocessors and keep the transistor budget within economical reason.
Today, we are able to make a prediction of where our technology may take us 15 years from now. Our best estimates, based upon past performance, existing technology and the laws of physics, show that the microprocessor of 2011 could look like this:
Transistors 1 Billion (435,000X the original 4004)
Die Size 1.8" ( about the size of a half dollar)
Frequency 10Ghz (4X the frequency of a microwave oven)
MIPS 100,000 (equivalent to 100 thousand VAX 1180s which were as large as refrigerators and were about 1MIPS)
The challenges that face us in delivering on this technology are to make our microprocessors faster, smaller and cheaper. Projected improvements in process and photolithographics technology (leading to a goal of .07micron technology) and the cost effectiveness of modern high volume microprocessor factories lead us to believe that the prediction of "Micro 2011" can indeed become a reality.
Beyond the Connected PC: The "War for Eyeballs"
The economics of our industry only work if we have large numbers of users demanding our technology. Consumers drive our industry now. They demand the highest performance and leading edge technologies. They are the early adopters of the latest in multimedia, Internet and communications. We need to be as relentless in our concern and efforts to grow the number of users and uses of our technology as we are in our efforts to develop and build the technology. Just as we are investing in technology for the long term, we need to target the new users of the future today. We need to look outside our own backyard for new users. We need to broaden our thinking about what it is we do and how we do it. There are things that future consumers of PCs want. Expectations of what their visual experience should be that have developed over the past 50 years. These expectations are clearly shaped by the television. We are in competition for these consumers, for their dollars and their leisure time. That competition is the TV. Consider that there are only about 1/3 as many PCs as TVs installed worldwide. While new PCs outship new TVs on a worldwide basis, we still have a long way
to go before we win this "war for eyeballs." In this war, "He who captures the most eyeballs wins." This rule applies to the Internet, and it applies to consumers' minds and the time they spend on their PCs. In our battle for eyeballs, users experience on the PC must not only meet the expectation levels set by TV viewing - it must exceed them.
Demonstrations: Multimedia PC Evolution and Visual Computing Evolution
Dr. Grove uses two demonstrations to introduce the concept of Visual Computing. The first demonstration is an evolutionary stair-step from what was considered state-of-the-art computing in 1992 to what is considered state of the art today and the performance of the video experience. The final step shows P6 processor-generation system with MMX™ technology and a video playing back in software only using MPEG2 and a DVD disk. The second demonstration shows the current state-of-the-art Pentium® Pro Processor system rendering a complex 3D database, and its current performance level. The second part of the demo simulates what the future of visual computing on the PC will look like by the year 2000. This final step shows visual computing at a level equivalent to the processing power of a 36- way Pentium Pro processor-based system today.
PC users wants more than ever before in the history of our industry. And the PC can meet their demand for real-world, interactive, three-dimensional and lifelike experiences. I am convinced that we can do it. After all, in just four years, we went from "postage stamp" video to broadcast quality. And the platform will continue to evolve from the connected PC of the mid-90s to the visual computing platform on the late 90s. The focus will be on improving the PC experience at the same relentless pace the PC industry has pushed all other boundaries of new technology. By the end of the decade the definition of a personal computer will broaden again to include interactive, lifelike experiences as part of the standard platform. To drive such an effort, we need to look at our business as more than simply the building and selling of personal computers. Our business is the delivery of information and lifelike interactive experiences. By redefining our views and the way we think of ourselves, we will be better equipped to deal with the future of our ever changing, ever-evolving industry and to make the next 25 years as productive, exciting and rewarding as our first 25.
Compaq, Compcore, Dolch, Evans&Sutherland, Informix Time Warner,and provided systems or engineering resources for the demonstrations in Dr. Grove's keynote. Please refer to each company's press materials for
* Other names and brands may be claimed as the property of others.