IBM uses nanosheets to build 5nm chip

Huge implications for the IoT

IBM has, apparently, done the impossible: laying the groundwork for a five nanometer (5 nm) chip - once 7 nm is solved, at least, which we're expecting next year.

In a project alongside Globalfoundries (Gloflo) and Samsung, the company used silicon nanosheets to build something even smaller than the current, seemingly unassailable (though of course it was always going to be eventually) 7nm chip, the smallest to use the fin field or FinFET design.

The new chip still uses extreme ultraviolet (EUV) lithography, but thanks to the nanosheets, it's a highly configurable, fine-tweaked chip that crams 30bn transistors into the same space as your pinky-nail.

IBM's interest is more in the area of Watson and other AI efforts, but there's no reason why the technique, if perfected, can't be used to make even more powerful yet small mobile devices.

The group claims that 5nm results could yield a 40 per cent performance boost at the same power, or a 75 per cent reduction in power at the same performance. What that could mean for things like IoT sensors could be quite phenomenal.

So-called GAAFETs could even be shrunk to 3nm, hypothetically. Beyond that, we're properly through the looking-glass but it would still extend Moore's Law beyond its current three-five year remaining lifespan.

Just as Concorde and the Apollo programme showed what was possible, but ultimately too expensive, it could well be that the new tech isn't cost-effective when it comes to balancing demand over potential saleability and we may never see any of this come to fruition.

IBM's technique is part of a miniature space-race with Intel who is still looking to keep Moore's Law alive as long as possible; but there's going to come a point in the next five years where it becomes physically impossible to shrink chips any further in their current form.

At that point, options may include a whole new technique that hasn't been invented yet, or something like quantum computing, which continues to be a subject of research and development for IBM.

Some people erroneously believe that there is some sort of conquering of physics to be done here, but Moore's Law isn't actually a law at all; it's a man-made construct that happens to have been true for a long time, but has no grounding in the laws of physics.