Aart de Geus, PhD, is chairman and chief executive officer of semiconductor-design and -manufacturing-software supplier Synopsys. He helped invent the Design Complier software tool and is a fellow at the IEEE.

What are the biggest challenges that your customers want you to solve?

I like to use the term 'techonomics' because there are technical challenges, and there are economic challenges, but they are on an intersection, which is: You can solve anything if you have infinite money, and, obviously, you don't. Our customers are rethinking everything they do right now around the notion of critical mass, which is not infinite money, but sufficient money to solve the problems, including doing it with others. You want new technologies, you can't afford it yourself, and [determining] who to team up with is a good example. You have issues in getting your design flow to end up with chips on schedule, so you'd better team up with somebody really proficient in the EDA category because all of these things interact.

Also, the gluing together of Frankenstein flows is [not working]. Frankenstein has run out of runway! More independent, best-in-class tools give worst-in-class results. These [approaches] are all forms of establishing critical mass around your issue. If you look at what the issues are on the design, it is very, very clear that two or three stand out.

One is the pairing of timing and power. Power is such a dominant issue, and there are some very cool escape values in terms of new architectures. But, fundamentally, it is a centre of gravity.

The second one is clearly everything that touches the word 'verification' because the verification problem is growing much more rapidly than Moore's Law for a variety of reasons. By the way, in verification, you have to throw in the embedded software because, ultimately, you have to understand about the physical aspects.

On the economic side, it is so much a market of 'winner takes all.' Therefore, if you're not on a big win, do you really want to continue? You may be better off searching for the next big win.

How is multicore impacting what customers want?

There are two very different angles to that [question]. One is multicore as a thing that influences how we write our software versus multicore as what are the things [we need to do] in design. In the first case, last year, we delivered quite a number of tools that are multicore/multithread, and some problems are better suited to split up. But, in general, there's no question that there will be more computation capabilities on a variety of multicore [products] - not necessarily heterogeneous ones - and that [trend] will continue. There will be all kinds of transition questions - and the transition may not be as fast as the capability, but we are utterly on top of it.

We have a centre of excellence within Synopsys dedicated to multi-fill-in-the blank techniques that helps all of our projects take advantage of this [ability]. We are engaged with the people providing the processors - Intel, AMD, or Nvidia - on what is coming.

On the other side, our insight into people designing these things clearly is a whole new world that is appearing on one hand, but it's not coming from the silicon side. The silicon side is just more of the same; it's just a different architecture. It's got to be on the embedded software, so I'm very interested in what that [software] is and the verification that is related to that.

Is the economic slowdown impacting Synopsys?

It's not an issue for us. But I do think it is an issue for a number of people because, whenever you have slowdowns or phase shifts in a market, it invariably means there are winners and losers. We've been predicting this [scenario] for quite a while on the very simple basis that, if you look at the economics of what's happening in semiconductors, they have grown nonlinearly, and that [growth] means that the price to play is substantial.

For example, if you want to develop the next technology node, a billion dollars is a good starting point. If you want to have a fab, $3 billion is a good starting point. If you want to build a chip, $10 million is a good starting point. You use those numbers, and you immediately conclude that not everybody can play. Now, that doesn't mean that there won't be a lot of economic activity; it means that the activity changes because, 'hey, why don't you and I share a fab?' That's called a foundry in some ways. Why don't I, instead of using five suppliers of something, get a much closer relationship with one that can then grow and give me better results than my gluing it all together.

When there is stress on a system, not everybody walks out equally happy, so the key is: Did one prepare and invest in the right aspects to be a winner under stress?

What is your plan for investing to win?

We've invested for a number of years based on five tenets. The first is: best-in-class products are necessary. Second, you need to have a complete offering because people are going to fewer suppliers. Third, the product needs to be integrated because products are no longer independent. The fourth tenet is 'roots down' because, with smaller geometries, the amount of physical issues bubbling up through design for manufacture is huge. And the fifth tenet is 'reach up' because more [functions are] embedded software, embedded intellectual-property blocks, and so forth.

By Ann Steffora Mutschler - EDN