Skip to main content

Beyond 32nm: the future of processors revealed

All the small things

When can we expect the results of all this trickery? Look no further than the International Technology Roadmap for Semiconductors (ITRS), which offers the industry's best guess. It predicts three more processes: 22nm, 16nm and 11nm. Chips using 22nm process are scheduled at around 2011. These will use similar techniques to the existing 32nm process.

It's predicted that this may be the limit for current planar designs however, and it might be necessary to build the gates vertically on what are termed 'fins'. Experimental 22nm SRAM chips are in the labs now. The 16nm process is expected before 2018, although Intel reckons it will be there by 2013. New problems will arise, including excessive quantum tunnelling.

This is where things get a bit weird as materials stop dancing to the rules of classical physics and Schrödinger's wave-equation pops up. Basically, it gets hard to stop the electrons breaching a barrier that's only a few nanometres thick no matter what material you use. This is also at the edge of commercial fabrication, nothing can currently be consistently and reliably made this small.

Toshiba has built a prototype memory module with 15nm lines, but such sizes are still lab experiments. Next stop on the ladder down is the 11nm process, predicted for 2022 by the ITRS, although, again, Intel is more buoyant and talks of 2015.

This is the expected limit of CMOS and may well mean silicon chips are no longer silicon. At this level dielectric thickness could be down to one atom, making it difficult to keep anything going where you want it to go.

It is also expected to be the end for conventional photolithography, etching and polishing methods. After 11nm the roadmap has yet to be drawn, it appears that this is as far as current technologies can take us. Possible ways of making even smaller and more powerful chips include three dimensional arrays, using nanowires and tubes, single electron devices, spinbased computing, photonics and any number of other weird and wonderful ideas.

Whichever proves the most commercial will win, business being business. Until then though, the good times will continue to roll, Moore is still right and chips will continue to shrink.