Despite the fact that it was based on an observation made more than four decades ago, Moore's Law continues to be as relevant today as it did in the early days of the semiconductor industry.
It's been the engine for change, the yardstick by which semiconductor manufacturers planned their future investments; the map by which semiconductor researchers plan the next process nodes.
Until the last years of the 1990s, transistor gate lengths decreased and performance increased without too much effort. For the last few years, however, the process shrink – the move to the next node – has proved to be more troublesome and engineers are turning to more exotic materials beyond 'plain silicon' in order to maintain the benefits expected to flow from adherence to Moore's Law.
ARM and IBM have just agreed to work on technology that scales to the 14nm node and both companies appreciate the challenges, because a whole raft of factors comes into play at those dimensions. Variability is a particular problem because there are so few atoms involved and less certainty about where they actually are. Anyone working on such advanced processes will inevitably be involved in the strange world of quantum mechanics, where terms such as Dirac cones, pseudospin, quantum tunnelling and excitons become part of the discussion.
Even if a 14nm process is developed, a lithography system will be needed that can create features at that scale – something that is proving an elusive target.
One small problem remains – what might replace silicon as the active component? Many researchers are asking this question right now, but the answer is not clear.
One thing is clear, however. The industry has spent too long reaping the benefits of scaling to give up now.
