CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

10 gigahertz or bust

Chuck Gwyn's consortium attempts to turbo-charge Moore's Law. Will it succeed?

If the chip industry is to continue boosting processor speed, it will have Chuck Gwyn to thank.

Charles Gwyn--he prefers to be called Chuck and he spells Gwyn with one N, thank you--is in charge of an ambitious project to build the technology tools that chip companies will need to create the next generation of microprocessors.

As general manager and program director of the Extreme Ultraviolet LLC consortium, this 60-something executive directs the research and development efforts of a 150-member team of engineers cherry-picked from the Lawrence Livermore and Sandia national laboratories, Intel, Advanced Micro Devices, Micron Technologies, and Infineon Technologies.

With the fastest PCs hovering around the 1-gigahertz mark, Gwyn's charge is to produce lithography tools that will lead to machine designs that can easily churn out microprocessors running at 10GHz or faster.

From his office at the Lawrence Livermore lab in Livermore, Calif., about 30 miles northeast of Santa Clara, Calif., Gwyn recently sat down for an interview to offer a project update.

You're directing a project to make the next generation of chips. How far along is it?
When we started this program in 1997 it was extremely risky. Now we would estimate the chances of success as being in the range of 90 percent. We still have a little bit of work to do, but we've demonstrated most of the elements of the technology.

The continued evolution of Moore's Law will make supercomputing affordable: A $1,500, 10GHz desktop computer would have the computing power of today's multimillion-dollar mainframes.What obstacles has your team encountered?
Initially there were a whole series of showstoppers. There were many who said we couldn't build the optics--that we couldn't produce a source that would generate sufficient EUV (Extreme Ultraviolet) flux--and then the whole system had to operate in a vacuum. It's been a lot of gradual progress, but I think we've made steady progress over the last three years. And we've got some very bright people working in the laboratory on the problem. We have a milestone this spring, which is the demonstration of this (alpha tool), the first full field-scanning tool.

How does EUV work?
Lithography is a process of transferring circuit geometries to the silicon surface. EUV lithography works in much the same way, except that we have to use a different wavelength of light and we need to use different lenses.

Is this stuff dangerous?
No, not really, because it's all enclosed. Of course the vacuum chamber is very thick steel, so none of the UV radiation can escape.

Why is EUV needed?
Basically, conventional lithography, which the industry has used for a number of years, won't be useful as we continue to shrink the geometries to a smaller process. We have to continue scaling the wavelength of light that is used. EUV is a very short wavelength of light, which allows us to bring very small geometries. The feature size you can print is proportional to the wavelength of light.

What will people do with chips that are 10GHz and faster?
If you have circuits that operate at 10GHz, which really are representative of the first technology that we will use EUV to build, you could imagine having real-time translators that would translate one language to another. Basically, you would have the equivalent of a supercomputer, found only in labs (right now), on your desktop. And, of course, there are all kinds of graphics and games that continue to use more and more processor speed. I think the industry has relied on Moore's Law (to date)...and this new technology will allow us to maintain Moore's Law for another decade.

Put this into everyday terms: How is what you're working on going to make a difference in peoples' lives?
Ultimately (in the 2005 time frame) we will manufacture microprocessors based on 30-nanometer transistors, and EUV lithography will be a critical component of that manufacturing process. The processors that result from this will operate at around 10GHz.

One of the most important advancements for end users could be in the area of human interfaces. Since you'll have a lot of horsepower, instead of typing or using the mouse, you can communicate with the computer by talking to it. Imagine this: You are doing your Christmas shopping online. Instead of spending hours browsing different sites, what if you could tell your computer: "Go online and find an XL green leather jacket between $200 and $400, buy the new Pearl Jam CD and a box of Godiva chocolates. Send the results of your search to my in-box and I'll choose what I want." That's the kind of thing you could do.

How about other "real world" examples?
A 10GHz processor could power a universal language translator very similar to a device used on "Star Trek." For example, an American tourist shopping at a market in Paris could speak English into a walkie-talkie-like device. The universal translator device would then automatically translate the tourist's question and audibly relay it in French to the shop owner.

The continued evolution of Moore's Law will make supercomputing affordable: A $1,500, 10GHz desktop computer would have the computing power of today's multimillion-dollar mainframes.

If you have circuits that operate at 10GHz, which really are representative of the first technology that we will use EUV to build, you could imagine having real-time translators that would translate one language to another.How much further down than 30 nanometers can you take current transistor technology before it won't work?
Intel and several university laboratories have demonstrated transistors well below 30 nanometers that will extend Moore's Law for several technology generations. Beyond conventional silicon devices, other devices will be needed, such as quantum and biomolecular devices, and alternate computing architectures will be required that continue the evolution of providing more computing power in smaller packages with lower power requirements.

What's the timeline for turning this research into real products?
EUV lithography is scheduled to support IC manufacturing at the 70-nanometer technology node. This requires preproduction tools for process development in late 2003 or early 2004 and production tools in 2005.

How is the technology that you folks work up here going to get disseminated to the rest of the chip industry?
The enabling capability will become available through production EUV lithography tools sold by the tool suppliers to the IC manufacturers. Each of the IC companies will need to develop the processes used in the fabrication facility. The EUV LLC companies will have an advantage of purchasing early tools and using the experience gained during the early technology development.

How important are advances in photo lithography to maintaining the relevance of Moore's Law?
Lithography is the enabling technology used in the IC fabrication facilities and is therefore key to manufacturing circuits with smaller feature sizes to maintain Moore's Law. EUV lithography will help maintain Moore's Law for at least another decade.

Is there a brick wall that you'll ultimately run up against?
We usually think of the end of a technology using today's known design and manufacturing methods. The end of conventional transistor structures could be envisioned when geometric sizes become so small that there are only a few atoms participating in the electronic conduction process. At that time alternative devices and manufacturing methods will be needed. But it is hard to say when that will be.

What do you think it might do in terms of lowering the cost of processor manufacturing--and therefore its impact on the cost of computing power for end users?
EUV lithography will support the continuing technology advancements described by Moore's Law, i.e., doubling the complexity of a circuit every 18 months while keeping the cost constant. This should allow industry to continue the historical trend experienced during the last 20 years of continuing to increase the computing power of desktop and laptop computers.

What would you want to do next?
I think I'll retire. This has been a challenge. The project is coming along very nicely. I believe we're going to finish it on schedule. I'll find something to do (after retirement).