For years, CPU designer Glenn Henry has been a Don Quixote of sorts, leading the charge for cheap processors with "good enough" performance. After working at IBM and then at Dell, he left the corporate world 11 years ago with enough money to be lazy for the rest of his life. So naturally, Henry did what any tech junkie would: raised funds for a startup in an intensely competitive business. Not long after, Centaur Technology Inc. was born, and Henry has been fighting the good fight ever since. Along the way, the Austin, Texas, company has passed through a couple of corporate org charts, ending up a subsidiary of Taiwan's Via Technologies Inc. As Henry explained it to EE Times' Mike Clendenin, computing for the masses is his mission.
EE Times: People are talking about cheap PCs, but do you think affordable phones for the masses will come first?
Glenn Henry: It's interesting that the phone is getting more functions, and the PC is getting more communications ability and becoming cheaper. But people need both. The killer application is the Internet. You need a decent display and decent input device, and a reasonable amount of computing power, if you are going to play any size video, or things like that. So the PC-based platform, but not in the conventional shape, is going to be the dominant platform. I don't think cell phones can do it. They will be cheap and are good for simple things, but they don't give you access to the real power of the Internet.
EET: Who will buy these cheap PCs?
Henry: A huge number of people can afford a $200 PC and want one, and don't have one today. I think you are going to see sub-$200 products appear that will be popular. We still haven't come close to hitting the market for sub-$200 PCs.
EET: What are the obstacles to building a sub-$200 system that doesn't skimp too much on specs?
Henry: The software is the single biggest cost element at that level, if you're considering Microsoft. But there are realistic ways around that if all you want to do is e-mail, a Web browser, etc. People are building many platforms using free software, such as Linux, that works quite well for that. Secondly, Microsoft has to do something about the price structure for this marketplace. Third, we all know that people steal software in countries where the people don't have a lot of money, so software in those countries may not be the most expensive thing. The next biggest barrier is the monitor.
EET: How about the total cost of ownership, such as servicing and Internet access or problems with stable electricity supply in some developing areas?
Henry: One of the barriers, even in rich countries, is ease of use of PCs. So that can become a real barrier when you are talking about countries that don't have a high rate of literacy or don't have a computer shop on every corner. I personally don't have any solutions for that. I'm the processor guy. But I think accessibility and ease of use are important, and there are technical solutions, which may raise the cost, to things like bad power supplies.
But I am less worried about that than I am whether the ordinary guy in Outer Mongolia will know what to do if he had a Windows machine sitting in front of him. That's why I like some of these closed systems, by which I mean . . . a box that boots up just to a Web browser, which means you can get access to the Web. That could be very appealing, and such a very low-cost device does exist.
EET: What's your profile of a person who might use these machines, and how would it improve their life?
Henry: If you look at China and India, there are a lot of poor people there, but there are also a lot of educated people there more than the total population of the U.S. By educated people, I mean people that have some job other than manual labor and have the equivalent of a high school education. Those people want access to information on the Web. They want e-mail and such basics. So I'm not talking about illiterate farmers, but people who are educated and live in poor countries.
EET: You say that you're just the processor guy. Do you get much time to get outside the lab and proselytize?
Henry: No, I don't get much time. It's not because I don't want to, it's because the story of Centaur is that we're a small team, very efficient, so we lock ourselves behind the doors and work hard. I give a lot of talks in the United States, and well, I was going to say that it falls on deaf ears, but in the last two or three years I have noticed quite a bit of change from the technical press and regular press that I talk to. People are beginning to realize that perhaps not everyone needs a 3-GHz computer, and maybe you don't need a new computer every two years. So I am seeing that story become more accepted in the U.S. But I don't really get a chance to talk to the real target audience of these cheaper PCs.
EET: You were talking about low power years before Intel did. Is this a case where the better technology doesn't always win?
Henry: I think the parts are really different. Intel parts are better than ours for some applications. We buy mass quantities of Pentium 4's for our chip simulations. On the other hand, our parts are better than Intel's for most typical applications. They are better because they are low power and not as costly. So these parts are really aimed at two different design points. If you want to do 3-D CAD design, then I don't recommend our chips. I would buy an Intel chip. But if you want to do e-mail, Web browsing, spreadsheets, databases or look at pictures, then use our chip, because it is better from a cost, power and performance perspective.
People have always asked me, since about 10 years ago when I was raising money for this company, about how we were going to compete with Intel. My real strategy is not to compete with Intel. If everyone in the world needs 3-GHz Pentium 4's, then we will fail. But I don't believe that is true.
EET: Now that the emphasis in the market is moving to low power and long battery life, is this the opportunity for Via's C3 and C7-M CPU architectures?
Henry: Yes, that is why we are so excited about the C7-M, because it does open up a new marketplace. Today, all of our parts are really sold into two marketplaces: the low-cost desktop and the embedded market, such as set-top boxes, media players and those things. We sell very little into classical PC mobile [applications]. The C7-M was specifically designed to be a good mobile part. Its power is better than a Pentium M, which Intel will tell you is the greatest mobile part in the world. We turned the megahertz up higher because megahertz is important in the industry, and the die size is particularly small, at 30 mm2, so the cost is very practical. We also have all the mobile features, such as thermal monitors, dynamic voltage and frequency. So this part opens up new opportunities for us.
EET: Why did you link up with Via and not a larger company that might have had more resources to help realize your vision of cheaper computing?
Henry: There are two factors. One is I don't think we ought to just be in the CPU business. You have to own a platform. If someone gave me a bunch of money to go off on my own to make processors, that's not a good business. So I was looking for a platform business, and one of the things about AMD, for instance, is they are not strong in platforms. They don't have their own chip sets, they don't have their own boards, etc. That's one of the great strengths of Via. We don't sell many of our processors by themselves. We sell them with a bag of parts, or the platform, and that allows us to do lots of optimization.
The other factor is the way we do business. Centaur is a very unique environment. We're designing competitive processors with a total design team of about 35 people. That's just ludicrously small by Intel or AMD standards. So it requires someone who believes in your approach and will send you the money and then leave you alone. That's what Via does. [Via CEO] Wen-chi Chen understood what I was talking about. Big companies try to mold you and what we do couldn't be done in Intel or IBM.
EET: You were at IBM for 20 years?
Henry: Twenty-one years. I was very close to the PC group, and when it invented the PC it was a very small, very talented group, sort of like my group [at Centaur]. It was 20 guys in a corner. Then all of a sudden, the PC became a runaway hit. And so IBM brought them into the mainstream and applied IBM organization principles and HR groups and watchers of watchers, and all of a sudden it was a 500-person group and IBM lost the PC business to leaner companies, like Dell. I believe that the kind of business we are going after means we have to keep our costs low, and you can't do that at a big company.
EET: Anything you would change about how you got your CPU to market?
Henry: We were originally owned by IDT. I think with IDT, too much energy went into trying to sell into the U.S. marketplace. It was natural because it was a marketplace we knew, but it just wasn't profitable. Now we sell about 50 percent of our parts into the U.S., but the marketing dollars that go with that are zero, so it's a good business. You'll never find an ad for us in the U.S. But if you go to Beijing, you'll find ads for Via, the mini-ITX [board], the processor.
EET: Should you have raised the clock frequency earlier?
Henry: We got it up as fast as fast as we could. I'm being honest with you. We started from scratch, just four guys in a garage. Two years later we shipped our first part, and the poor thing only ran at 167 MHz. But it sold. These things are hard to do, so, yeah, if I could wave a magic wand and say "make the megahertz higher," then I'd do it. But there weren't any decisions or mistakes that were made that prevented us from pushing it up higher. It was just a matter of developing a team, the methodology and the tools.
These parts are custom-designed, so we had to develop our own tools. Very few people do custom chips anymore, other than processor guys, and X86 verification is the greatest secret of all. No one talks about it. Those are all methodologies that we've had to develop along the way, by ourselves.
July 26, 1942, Berkeley, Calif.
California State University, Hayward, BS in mathematics, 1966; MS, 1967
IBM Corp., 1967-1988
- IBM Fellow
Dell Computer Corp., 1988-1994
- Senior vice president, Product Group, and chief technology officer
Centaur Technology Inc., 1995-present
- Member of IEEE and ACM
- Holds a commercial pilot's license and flies a Cirrus Design SR22 aircraft