What will the future of computing look like? Intels Pat Gelsinger is spending about $4.2 billion this year to get some answers. Gelsinger, 40, is an Intel lifer who started at the company in 1979 as a teen-ager fresh from technical school. He was recently named chief technology officer — the first CTO Intel has ever had. In that role, Gelsinger will direct the R&D activities of 6,000 people in 80 labs around the world. He spoke with Matrix Editor Todd Spangler.
What are some new areas of Intels research?
More and more, were doing higher-level innovations in software — things like peer-to-peer services, Web services and proactive computing. We have a lot going on in the standardization of the Web services infrastructure. Beyond that, were beginning work on proactive computing research, which is the next big step.
What is “proactive computing”?
Today, computing is essentially reactive: You tell your computer to do something. Its responding to human action. With proactive computing, you have intelligence in the network itself that is doing things proactively on your behalf. We recently did a demonstration where we had about 1,000 nodes in a self-configuring network. The network itself was smart enough to tell adjacent nodes in the network how to configure themselves.
On a separate topic, whats the limit on how small the circuits on a microchip can get?
What we demonstrated [earlier this year] was the first 20-nanometer [20 billionths of a meter] transistor, and we showed the basic operational transistor that wed be putting into production in about 10 years. Based on that, wed be able to produce a single piece of silicon that had over a billion transistors on it and would be able to operate in the 30-gigahertz range.
One of the challenges wed have in making such a device is power. We sort of blindly went from sub-1 watt to almost 100 watts in our current products. Now, if you scale that forward over the next 10 years, you get to products that are well over 1,000 watts. So weve had to attack this new problem of power and start to play out some directions we as an industry need to go to accomplish that.
When you have that 30-GHz chip, what are you going to do with it?
You know, thats a whole lot of honking horsepower. [Laughs.] Weve been looking at some areas that could take advantage of such tremendous horsepower. If you think about looking at somebody when they talk to you, youre getting not just audible recognition; youre getting visual recognition as well. In normal conversation, its the visual cues that give you that additional recognition to go from 90 [percent] to 95 percent recognition to 99 percent recognition of what they actually say. So one of the areas were researching is combining vision recognition with speech recognition. Vision recognition by itself is a many-gigahertz problem. Then you combine that with speech recognition — which is already a multigigahertz problem — and you start to consume vast amounts of computing resources.
How much research is Intel devoting to quantum computing [which applies principles of quantum mechanics to massively parallel processing]?
None.
Why not?
We have a research horizon of things we think are about 10 years out, and quantum computing is 20 [years] or 30 years out. What we do is partner with people who are working in those areas. Recently, we started “lablets,” where well actually put labs in or adjacent to a university, and fund some people at the university. So we have a very rich model of interacting with research institutions, and that gives us that 20-[year] to 30-year window for things that we dont want to put our own nickel directly into, because we dont think its quite ready for us to start really investing in.
At the end of the day, what really matters isnt the Nobel Prize or that next big research breakthrough people are dreaming about, but the companies that can take those breakthroughs and actually make products that change the world.