Three reasons give designers an edge over manufacturers in product differentiation. First, manufacturers make products for many competing customers. For instance, foundries make semiconductors for rival semiconductor vendors. In this scenario, the manufacturer cannot provide differentiation, leaving greater responsibility than before on the designer to plan product differentiation. Second, with vendors choosing to wean themselves away from manufacturing in varying degrees, R&D revenues that would otherwise have gone into manufacturing are now going into designs, giving the designer greater leverage to work on product differentiation. Third, most vendors aim at worldwide markets varying widely in economies, cost structures, attractive prices, etc. The same design will not click in all markets. The onus has fallen on the designer to build designs to make product differentiation answer to regional demands.
MAKIMOTO’S LAW When thinking about product differentiation, designers would do well to keep in mind where the industry is placed on Makimoto’s wave. According to Makimoto’s Law, product differentiation passes through a cycle of standardization and customization. In his words, “When large numbers of new technologies, devices, architectures, and software appear, the semiconductor industry as a whole moves towards standardization. Then aspects appear that suppress this motion, this large swinging of the pendulum. These are the need for product differentiation and added value, and the imbalance between supply and demand. In the other direction, progress in design automation and advances in technologies such as CAM and CAT occur, and the semiconductor industry then shifts to customization. . . . When the whole semiconductor arena becomes oriented towards customization, there then appears reverse trends towards early market entry, cost reductions, and more efficient operation. If the whole industry moves towards standardization, then a push towards customization rises, and when the industry has moved towards customization, a force for standardization pushes back. As seen from a macro viewpoint, the semiconductor industry can be said to repeat alternate phases of standardization and customization.”
In the 1960s, the standard product was in logic, TTL, and memory. The standard product gave way to custom designs for calculators and other emerging consumer products. Custom design gave way to standard design in the 1970s and early 1980s, with the advent of the microprocessor, a general-purpose product used in many applications. In the mid-1980s ASICs stormed the design industry and gave designers full-custom capabilities rather than general.
Makimoto says that the next five years will see field programmability becoming a key tool in the designer’s hands. After that the industry will swing to customization again, and system LSI technologies, such as SoC and SiP, will come to the forefront.
KEY TECHNOLOGIES “The key technologies at that time,” Makimoto says, “will be maskless technologies that reduce the increasing mask costs, superconnector technologies that can support complex multi-level wiring technologies, and e-businesses that can take full advantage of networks.”
I think there are some areas that will continue to move towards full custom design. Applications with very low power such as handhelds, and applications involving analog and RF circuitry place more premium on technological feasibility and are likely to be full-custom. Also, high-volume applications will find custom-design more cost-effective than programmable.
Perhaps the industry might think differently of what exactly a custom design is. Today a custom design is a design not including programmable. But increasingly designers say that a chip with programmable logic programmed to do something unique is also custom design. The customization is taking place only at a particular level. This might lead to a situation wherein custom work done directly in the silicon will be less than programmable elements on the silicon.
