Viewpoint

Real Men Have Layout IC design will increasingly depend on migrating old intellectual property to newer technologies, so companies must own their layouts. by Hein van der Wildt

The EDA industry is going through a time of tremendous consolidation, and the reasons are obvious. Previously seen as having a boring business model with little or no growth potential, the industry has recently expanded because some brilliant minds have discovered that we can provide services to the semiconductor industry. Although providing services has nothing to do with EDA, it's a way to increase revenues.

The EDA industry is like the automobile industry. Auto makers typically spend a lot of time adding features to a trusty model, rather than overhauling the engine to create a newer, faster car. Similarly, for a long time, EDA management has been preoccupied with product maintenance and has thus failed to develop the faster, newer products designers need. We put far too much emphasis on adding features to existing products, instead of revamping the entire product. So we got stuck with difficult-to-maintain tools that, by now, are so complex you need an army of engineers to support their use. In essence, when you buy an EDA tool today, you get a '97 body with a '57 engine.

The manufacturing industry, on the other hand, is growing rapidly. Very deep submicron technology--process geometries down to 0.10 µm--is no longer a goal of the distant future. Indeed, many influential semiconductor manufacturers are moving in this direction now. But the designs are lagging, and the fabs aren't being loaded fast enough. A $2 billion fab with a life of about three years depreciates in value at an amazingly expensive rate--up to several million dollars each day--so that a day lost in fab loading can be a disaster.

Where are the designs for the next-generation, $2 billion fabs coming from? Internal design groups can't and won't ever be able to leverage the new technologies to generate new designs. The tools are inadequate, the manpower is insufficient, and the time-to-market requirements are too stringent.

The answer? We have to reuse existing designs and migrate them to the new technologies. Of course, the EDA industry can and should also provide services to assist the semiconductor industry, but at the same time, it should provide the tools to migrate and re-lay out existing intellectual property (IP). To reuse designs, the semiconductor and electronic system companies need to own their own layout.

Very deep submicron design issues can't be solved at the behavioral level; they have to be solved at the physical. No matter how good you are at designing, no matter when your design is DRC-, ERC-, or LVS-correct, if you don't apply wire spacing, correct for optical proximity, and address crosstalk issues, your yield will be unacceptably low. Migration needs to be driven by timing and power dissipation, combined with formal verification and device sizing, so designers need to get their heads out of the abstract high-level clouds and put their feet firmly on the ground.

First of all, companies are beginning to take inventory of their existing IP. Instead of redesigning it, they're categorizing what exists and forcing their design communities to use it. Second, companies need tools that enable them to migrate, re-lay out, and optimize their existing designs at the layout level, allowing them to create new IP that satisfies timing and power dissipation requirements and giving them new, optimized designs that take advantage of new technologies.

It's rumored that "real men have fabs." That may be true, but I assert that the slogan for successful design companies in the next century should be "real men have layout," or rather "real men own layout." More and more, design houses are adopting that strategy. They develop, own, and migrate their own libraries; own their own IP; and consequently cash in on tremendous opportunities. If there was ever a way to provide real benefits to the semiconductor industry, here it is.

Hein van der Wildt is president and CEO of Sagantec, Inc. in Milpitas, Calif. He previously managed the Dutch-Swiss-U.S. multinational ENBI Corp., establishing U.S. sales and manufacturing operations for high-precision components for Canon, Hewlett-Packard, IBM, and Xerox. Prior to that, he was European managing director of Xerox Computer Services and audit manager for the Rank Xerox operational auditing department. He began his career as an electronics engineer.

To voice an opinion on this or any Integrated System Design article, please email your message to miker@isdmag.com.

integrated system design  June 1998