If Moore’s Law ‘hits the wall’ in 2020, what comes next?
1 min read
We have become used to the semiconductor manufacturing sector meeting the requirements of Moore's Law; every 18 months, they produce chips with twice as many transistors per unit area.
Over the last few years, however, it has become more apparent that Moore's Law will 'hit the wall' at some time in the near future. The reason is simply the Laws of Physics. While there have been informal predictions about when, nobody has gone on the record – until now. Dr Robert Colwell, director of microsystems technology for DARPA – the US defence technology organisation – told a recent conference it could be as early as 2020.
It's not a prediction that will come as a great shock to the industry, but how great will be the impact of manufacturers not being able to continue to scale their devices?
Pushing down the technology curve has enabled devices to operate more quickly, to consume less power and, depending on your requirements, either make smaller versions of an existing part or put more functionality into a given area. Will the world be any the worse off if scaling comes to a full stop at the 5nm node? After all, many devices are still made on what are known as 'legacy processes'.
But Moore's Law is concerned only with linear dimensions; when Gordon Moore framed his hypothesis in the 1960s, nobody thought about using the Z direction. Recent developments in the flash memory business probably give a good pointer as to where chips will go.
Rather than going through an elaborate back grinding process to produce thin dice, we will be stacking them in a package. It is a process that allows many layers – Samsung says up to 24 – to be created on one die. Because the chip is being built upwards, pressure is relieved on the need to squeeze the X and Y dimensions. Micron, another flash manufacturer exploring 3d memory, says it can relax these dimensions significantly.
Perhaps building upwards, rather than shrinking planar dimensions, will keep the wheel turning and allow time for more exotic approaches to be developed.