IC designers today are faced with demands for greater up-front planning, more back-end testing, power economies, testability and tool functionality. What platforms, tools and services will leaders invest in and why? How will open-source architectures and "virtual" design capabilities change electronic design automation methodologies? And which EDA platforms hold the most promise for ensuring greater price/performance advantages?

The many choices that EDA engineers must consider going forward offer us a futuristic view of next-generation EDA environments. Factors that face electronic designers include global competition, time-to-market, engineering and support resources, product complexity and ever-decreasing design cycles. But cost continues to be the pivotal concern.

In an analysis of the relationship between EDA expenditures and the competitive positioning of IC vendors undertaken in November 2002, International Business Strategies determined that the typical IC development of a 90-namometer IC design is expected to cost roughly $54 million over the life of the product. Slightly more than $24 million will be dedicated to hardware, with the rest going to software. Consulting fees and other maintenance services will likely add an additional 7 percent. This study also calculated that the percentage of year-to-year EDA R&D spending over revenue is highest among leading IC vendors.

In what are these leaders investing-and why? Linux doubled its penetration in the EDA market in 2002, overtaking RISC-based Solaris as the main secondary EDA platform. And Linux is predicted to be the main primary platform within 12 to 18 months. The market share for HP-UX and Windows NT, according to a platform study by Synopsys Inc., is negligible, while Linux platforms increased over the two prior years.

Industry experts examine trends like this to gauge the shifts in competitive price and performance points that these EDA systems are able to deliver. Particularly as EDA moves from 32-bit to 64-bit systems, open-architecture platforms like Linux have inherent advantages in their quest for dominance. Leading EDA vendors like Cadence, Mentor Graphics, Synopsys and leading platform suppliers such as IBM are working together to port their tools to Linux-and for good reason.

ASIC complexity, depicted in a 2003 study conducted by Gartner/Dataquest, is on the rise-even as the number of design starts are projected to decline through 2006. The emphasis here will be on exploiting the latest design tools needed to effectively run up-front planning, conduct more back-end testing, restrict and control leakage power, ensure synchronization and incorporate testability.

Tools that commonly pose the most problems include IC place and route, design for test, logic synthesis and functional verification. Because the typical 90-nm mask set costs more than $1 million and cycle times are slimming from 24 to 12 months, CAD/IT teams will dominate the tool-buying decisions. More than 85 percent of all tool purchases will be decided by CAD/IT departments, typically in concurrence with design engineers and other engineering support staffs.

How will open-source architectures and "virtual" design capabilities change EDA methodologies? New Web, grid and wireless protocols will require more design houses to develop even more scalable and highly integrated solutions that bridge today's boundaries. In a revealing report produced by Forrester Research titled "The Linux Tipping Point," design companies were asked whether their use of Linux would increase during the next two years. Nearly three-quarters of the respondents said it would.

The study did not convey how many of these companies operated their entire EDA programs at single locations. But it is reasonable to assume that whenever a company's engineering talent resides across international borders, open-source platforms built on grid-computing architectures are becoming the solution of choice. That's because distributed (or cluster) computing strives to pool the processing resources of computers and workstations in a series, whereas grid computing takes this concept one step further.

Bear in mind that an Intel-based system running Linux can be a high-performance solution and also very cost-effective. In a multinode compute farm implementation, it is generally easier to integrate new platforms. Given the price/performance gain achievable in such a system when compared with traditional RISC-based platforms-from which many EDA vendors are shying away-the focus is growing on technical assessments that deliver value today and promise more ROI tomorrow.

Grid computing (distributed computing over a network using open standards to enable heterogeneous operation) offers far more detailed scheduling, higher levels of service and fully functional distributed control. It enables efficient and more dynamic resource provisioning. The result is a seamless and virtual data access experience across the entire enterprise.

Utility computing will proliferate as well. Its attributes include high-quality service, superb efficiency, simplicity, improved resiliency and increased productivity. In a utility-computing environment, server infrastructures are a major source of corporate inefficiencies. Server utilization rates are typically lower as well: running between 8 percent and 20 percent of maximum. Utility computing provides an adaptive, integrated and reliable infrastructure that delivers on-demand services for on-demand business operations.

http://www.eet.com