High speed: not the only way to exploit the intrinsic computational power of silicon

Abstract

A seemingly insatiable demand for bandwidth is induced by continuing progress in silicon technology. This progress results in the ongoing emergence of computationally powerful new digital systems which, in turn, challenge the progress in technology. In this way a positive-feedback loop has come to exist, resulting in the need for handling and communicating ever more data. Exploitation of the computational power of silicon is the key to satisfying the perpetual need for more bandwidth. However, important as it may be, the computational power of silicon is mainly an intuitive notion. It stems from attempts to characterize the capabilities of a piece of silicon by the benchmark approaches commonly applied to the best-known of all silicon devices: the microprocessor. Clock speed of a microprocessor often is regarded as the decisive performance indicator for computational power and many people believe that increasing speed is the ultimate art of digital design. For this reason the clock speed of a microprocessor has become an important weapon in aggressive marketing and advertising aimed at convincing consumers to buy a particular new piece of equipment. But, is it really true that the computational power of silicon depends, above all, on the speed of the microprocessor? And if the answer here is no, where does the strength of a piece of silicon really come from? What will be the benefits for the average consumer of endless boosts of computational power and bandwidth? Will he/she not be overwhelmed by what clearly appear to be exponential developments obeying Moore's law? An attempt is made to answer some of these questions.