News and New Products

Cadence's beat

By Suzanne Deffree, Managing Editor, News -- Electronic News, 12/18/2007

Electronic News/EDN recently sat down with Mike Fister, president and CEO of Cadence Design Systems Inc. and the former senior VP and general manager of Intel Corp.'s enterprise platforms group, to discuss the EDA industry, work across the capital-equipment value chain, and life after Intel. What follows are excerpts of that conversation.

Q: How's business?
Fister: What's hot is still hot. Consumer electronics has always boded well for us because of the mixed-signal strength. It's fascinating to see how much more integration is going in here and the tremendous pressure on the semiconductor cost. Some of that semiconductor innovation isn't highly valued. It sounds like the old days at Intel.

Q: Still drawing a lot on your Intel experience?
Fister: I always have. It's not only one of the best known companies in the world—I'm not biased, of course—but I think its always been a protagonist for large volume, how economics drives that cycle, how innovation drives that cycle, and where the foundation of innovation starts. There's one other point. Go back to when I was a young guy and working on software to amortize volume in the semiconductor industry. The application of that is incredibly tricky. When you do that, the validation problem becomes extraordinary complex. You become more platform-driven, thinking about structure architecture, and software becomes an integral part of it. That's one of the reasons we drove an adjacency around verification. When I came in [to Cadence], I had done that for Intel in 1995. Every one of our customers builds a system-on-a-chip as a platform now. Financial and fundamental impacts are hiring software and verification experts at record rates, but also trying to anticipate that is mind boggling. It's one of the reasons we are at a "perfect storm" for automation. You have that complexity, time-to-market, and incredible cost pressure. Automation has always been one of the things we use to chart progress there. We work hard every day to prove that and achieve reasonable value for it. You don't want to do it for nothing. 

Q: How long have you been at Cadence now?
Fister: Three and a half years.
 
Q: How's it differing from your life at Intel?
Fister: In technology companies, technology people are what make the thing go. We're lucky to have a lot of talented people in the company, and we've brought in more semiconductor guys. That was to be able to have that experience base to rest on so you have an idea of where the roadmap is going to go. You don't want to do this and hope you are doing it right. Hope is a lousy strategy. In those respects, it's the people side.
It's interesting to me because I see a much broader section of the world. Certainly operating across the world with customers, but now we are fundamentally into the educational ecosystem and the business dominance of it. We just opened an office in Zelenograd [Russia]. We go for locality where the customers are. China continues to grow nicely, and India. The emerging geo theme is something we have a lot of passion about as we did even back then. So maybe there isn't a lot new there. Fundamentally, what we do is just like semiconductors. If you have a microprocessor, this is how much it is and most people use this much, that's why we stratify, to get that investment. The thing that's difficult is to really make some of these companies say, "Hey, this is fundamentally valuable to what I do. How do we share in that risk and reward?" Maybe it's not the same in semiconductors, because they are highly commoditized, or maybe it was. The breadth of companies from fabless startups in China or India to the biggest IBMs of the world and the changes to the ecosystem are very interesting. I think we have an opportunity that's extremely explosive in the growth dynamic that we have and the value dynamic that we have. I don't think traditional semiconductors have much of that.

Q: How big of a challenge has it been for you to go out and convince some companies in the industry of the value of EDA?
Fister: If you get to the right person, it's not. I came into the industry valuing three things that I think some people thought were marketing: time-to-market, complexity management, and productivity. Those are absolutely the objective measures that guys who run semiconductor programs use. If you deal exclusively with an engineer—who may have been a genesis of a tool and you are talking about their potential livelihood—some of those conversations are extremely down into the detail. They are benchmarking-crazed and, you know, you can manipulate those kinds of benchmarks. Then it becomes very much a history of EDA, point by point by point. When you get to that person who you can show those measures of objectivity, it becomes nondebatable. If you can demonstrate value, you get some of that value back. That's one thought. The second thought is, some people move faster than others. It takes a great person, whether an engineer or manager, to try something different or legislating discipline than it does for someone to do the status quo. Some people will say, "It's the devil that I know." And I get that. But leaders lead. I think back to my alma mater and we would make calculated guesses and drive discipline in methodology as strongly as we would in building the product. The way you do it is as important as what you do, sometimes, because that's how you get the result, especially when you begin to compound these co-design elements of software and validation and marketing value. 

Q: Who are the first movers?
Fister: Some of the people that are first movers on the consolidations that we have partnered with are a pleasure to work with and there are lots of them. We saw another one last quarter with the Philips spin-out of NXP. To me, companies like Hitachi are testimonies. These are computer companies, not just semiconductor companies. I am saying that very carefully. They are the role model of "platform": system and software engineers to complement implementation, either on the semiconductor or printed-circuit side, for total product. They see very nicely what we are doing up the stack and the seamlessness of the predictability of how you implement it and eventually how you manufacture it. Israel is on fire with so many startups doing crazy challenging devices, fortunately every one of them has some element of mixed-signal or extremely high-performance digital. And a lot of those companies are buying with us end-to-end. It's not unlike what happened in the PC industry. You had large vertical companies and people like [Michael] Dell came in with methods that seemed disruptive at the time, concentrating on end-to-end including the delivery dynamic, and innovated some, and they put more on reliance with partners in semiconductors. It happens in every industry. 

Q: What's going on in the DFM space? We've been hearing about DFM for years and Cadence recently made some acquisitions there, most recently Clear Shape in August.
Fister: Design for manufacturability is a little bit of a confused category, as far as I'm concerned. A lot of it is from the genesis of any use scale, kind of a physical rules sign-off set that's a post-process design process. And it's also in many respects dangerous because when you do something like proximity correction and modify the design, imagine the design guy. He looks at it and says "I gave my design and signed it off in a perfectly good way. You then monkeyed with it. If it doesn't work, my design was fine when I gave it to you, and you broke it." The other thing is, complexity has gone up. There are supercomputers trying to do the conversions for best-known methods applied across lithography and/or the convergence of that can take months. It didn't take a genius to try to intercept that.

Q: It's been reported that Cadence is working with Applied Materials.
Fister: We are working with equipment manufacturers, as well as the foundries. I often say we are a glue, connecting that virtual path between equipment manufacturers and foundries, internal or external. EDA was one of the first things that disaggregated out of companies, then manufacturing and foundries, and now process development. It's important to be working with those constituencies not only for the innovation that they are doing but I think sometimes, in history, innovation in the ecosystem was prominent with creative capability and they would look for a way to automate and measure that. It's something that makes it more complex for us, but the biggest companies, the IBMs, can do it themselves, the big fabless guys can do that, but not everyone can touch every aspect of the ecosystem. It puts some onus on us to do that and be very good at it.

Q: How do you see equipment, litho, and EDA working together in the near term?
Fister: Increasingly important for all of us, just harder. We see these breakthroughs now on CMP techniques as a way to control or influence the predictability of the metallization. We've been doing techniques like that integrated into the tool for a long time. The results that you get are just mind-boggling. It's not you do something then tell me and I figure out how to make it mind-boggling, it's what should we do together to make it mind-boggling as the process. Increasingly, that's the point. You know, you can't do that with every partner. For a long time there was a bit of a "chicken and the egg" as to, how do I work with anybody if I'm part of the ecosystem? You just see people nominally making choices. And you see the momentum of leadership. When I came into the semiconductor industry, everyone was in memory. For a while there were logic blocks with little pieces of memory. Now there are huge memories with little logic blocks because memory was the architecture we would use for interconnection.

Q: What other avenues do you see for growth?
Fister: I'm excited about verification as an element. It's a second kind of adjacency that allows us to do two things: It allows us to reach up and touch system designers and software designers with the techniques we are doing. A lot of the techniques we are using as a logic designer to functionally validate our ideas, we're establishing now to software people. They use the same model to develop the application in parallel with software development. Usually it is for pieces of a chip, but I think we are going to see it exploding with big farms with those kinds of emulators as kind of an example of the next server farm for coverage dynamic. I think this breakthrough around co-design with the software we're on, and you can do it model, verification or validation, post. The way I've described it is from the core of EDA, implementation, stepping up stairways, touching these things.
With all of the core integration of the die, you trip over one of the hardest problems that people really have solved forever, it's just distributing the application across those pieces. You start off doing that kind of way where you allocate jobs, then as the interconnection or communication rate picks up, it becomes complicated. The best way we'll attack that is by using verification then validation, pre and post, to be able to get after some of those and test the infinite universe people are doing. Then we'll learn things about elements that are predictable and we'll apply formal mathematical methods to take those and make them better. That's the history of automation once again. We lay the foundation in those formal methods. There's a logic to how we are applying that; it's not just random that we are doing that stuff. We are guiding those roadmaps for exactly that reason. So I'm excited about that.
And it leads you to touch verification IP and design IP. And also then the virtual platforms. When you are doing this architectural stuff, what are you doing? Trying to make tradeoffs. The number of big blocks you can have on chips is awe inspiring.

Q: You haven't served as chairman of the EDA Consortium (EDAC) yet. Would you be interested in that?
Fister: It's a pleasure to participate in it. I think we have some ability to help influence the momentum of the EDA industry as a leader in the industry, as many of my other colleagues on the board do. We've tried to take some best known methods and extrapolate them to other companies, share learnings of operating in these emerging geographies. For what it is, that's fine, and then there are things it's probably not as influential in doing. Standards converging has never been much of a focus. I kind of go for whatever we can do.

Q: It doesn't sound like you would want to take that on.
Fister: I don't know. No one's come and asked me yet. I'm active in whatever I'm doing, whether I'm the leader of the band or a participant.