Summing up 2003, and looking ahead

As we reach the year-end, my thoughts go back to the last days of 2002, when the industry was seesawing with hopes of a turnaround. The hopes were belied. In 2003 end, we are better off. We are more confident than before of a recovery from the devastating recession. Though layoffs continue, promising figures emerge. These figures are of capacity utilization and growth, and not prediction. After the battering prediction figures took during recession, they will take time to acquire their magic again.

To be a part of the magical world has been in humankind's imagination since long-whether in the enchanting tale of Cinderella's fairy godmother turning a pumpkin into a carriage, or in Aladdin discovering a lamp with a genie who makes every wish come true, or in J.K. Rowling's Harry Potter with his bewitched companions from the Hogwarts School of Witchcraft and Wizardry. In 2003 the electronic design industry firmly launched itself into such a magical world. "Designs being right first-time" is the magical concept driving design engineers today. This concept must turn into reality. In every human endeavor, first-time successes have been a remote exception. But in the design industry, unless first-time successes become the rule, economic benefits from nanometrics will not materialize.

Notwithstanding all efforts for first-time successes, in 2003, 61 percent of ICs and ASICs are being redesigned and returned to the fab, according to Aart de Geus, chairman and CEO of Synopsis. Functional logic errors are the largest, constituting 43 percent of errors, followed by analog tuning errors, at 20 percent. The year saw chip designers, chipmakers and EDA toolmakers coming together like never before to achieve first-time successes. The challenge passes on into 2004. In 2002 EDA tool inadequacy was said to be the reason for high design failures; but now it is known that EDA tools are only a part of the problem. An entirely new design paradigm is required, but nobody knows what it will be.

SoC has reached the Hamletian dilemma "To be or not to be." SoC is the ideal towards which the industry is moving with the speed of Moore's Law. But as the industry stepped into 90nm node in 2003, SoC seems to have hit a wall. The concept of mixed signal integration does not work efficiently at this node, and chipmakers are forced to make trade-offs. Two major mixed-signal players, Analog Devices and National Semi-conductors, are changing path from SoC towards multiple chips in a single package or in different packages. Is this the harbinger of a new direction away from SoC? It is too early to say so, with most companies still pursuing SoC. However, the challenge to SoC rose to such an extent in 2003 that the Semico summit thought it fit to discuss whether SoC was "dead." The next year should see corroboration or futility of the SoC model? Integrating everything on a chip may no more be a dream worth pursuing.

Design costs and delays, which started taking precedence over production costs and yields with 130nm node, continued to be a major concern in 2003. The problem rolls over into 2004. With economic considerations superseding technological considerat-ions in semiconductor industry, for the first time the question arose on the relevance of Moore's Law. Most still believe that Moore's Law has life left, but the number of opponents to Moore's Law is growing. The next year should see a tug-of-war between the two groups, and perhaps the final judgment on Moore's Law validity.

The major technological challenge is on the power front. The problem is bigger than thought earlier. At 90nm node, power leakage is so high that the concept of "on" and "off", forming the basis of digital engineering, does not seem to be working reliably. What is the solution? Some suggest multiple power domains? Some, prioritizing functions? And some, using older-generation chips? Does technology move in a cycle, like life?

The uncertainties have thwarted the rush into nanometrics, and 180nm is set to stay longer than expected.

You can reach Kirtimaya Varma at kirti.varma@rbi-asia.com