Current designs for commercial microprocessors and the ways they are arranged to make a computer are more than 20 years old and well past their sell-by date. Their inefficiencies are preventing the industry making the most of developments in semiconductor technology to increase performance.

Scientists in Texas have embarked on a promising attempt to overcome these bottlenecks - for instance, in wiring and memory access - and to develop a superior design for mainstream computing. International Business Machines' Austin laboratory is working on a research project with the University of Texas to produce a supercomputer-on-a-chip. When scaled up, the architecture could handle as many as a 1,000bn mathematical operations per second on one piece of silicon.