New growth path for transistors

Early this year both Intel and IBM announced perhaps the most significant transistor overhaul technology in the last 50 years. The two companies separately declared that they had found a way to reduce energy loss in microchip transistors scaling down to the atomic scale. The new technology is now said to be ready, with products likely to hit the market soon.

What will be the impact of the new technology? For the last four decades and more, Moore’s Law has provided a fairly accurate path to the growth of the number of transistors on a chip. However, at 65nm and lower nodes designers have increasingly faced the problem of leakage current. Silicon has worn so thin that an increasing amount of current seeps through, causing not only electricity wastage but also heat dissipation problem. Designers have been grappling with this problem even in nodes higher than 65nm. However, no satisfactory solutions could be found. Intel and IBM have created new alloys for building insulating walls and switching gates that can take care of the problem. Intel claims the new materials have boosted transistor performance by 20 percent. IBM, which has done the project in coordination with AMD, Sony, and Toshiba, has yet to announce quantitative improvements.

STRETCHING MOORE’S LAW
Intel confirms that chips using the new materials will go into mass production in H2 07. The next-gen Intel Core 2 Duo, Intel Core 2 Quad, and Xeon family multi-core processors will use these new materials. Especially interesting is Intel’s Penryn family of 45nm processors with 400 million transistors for dual-core, and 800 million for quad-core, which includes the new micro-architecture features for greater performance and power management capabilities, and also higher core speeds and larger caches. IBM will not sell the chips by themselves; it will ship servers based on these chips next year. AMD has not yet disclosed when it will use the new materials. Apparently Intel is ahead in implementing the new architecture.

Designers agree that the new developments have given a new lease of life to Moore’s Law, which just one year ago was said to hit a wall. The chip will continue to advance on all the three main metrics—performance, transistor density, and power consumption—in conformity with Moore’s Law. Indeed, some designers claim that Moore’s Law is now poised to provide the transistor growth map well into the next decade.

Moore’s Law was said to be at the end of the road for two reasons: first, technological feasibility, with leakage current and power dissipation being the most serious challenges; and second, economic feasibility—the ability to get more power and performance at lower cost. The new materials could provide technological feasibility for the continuation of the law. But what about economic feasibility?

DOUBLING PROCESSING POWER
I think Moore’s Law should be seen from two angles. One is doubling of number of transistors on a chip every 18 months. Another is making available to the consumer double the processing power at half the cost every 18 months. We generally look at the law from the first angle. I think we should now look from the second. The consumer does not care for the number of transistors on a chip. What he cares for is processing power and cost. For doubling the processing power at half the cost, designers don’t have to double the number of transistors, but can achieve their goal through appropriate hardware and software designs, and packages. This provides designers not only with versatility, as there are available to him various areas to work on, but also economies, because the cost of these designs is enormously lower than that for increasing the number of transistors on a chip.