Now that 7nm hardware is closer to the market, its time to consider whether 5nm can be made practical enough for production. At the smaller sizes its helpful to consider the picometer (pm) which is 1000 times smaller which is more helpful for the smaller nodes. 7nm becomes 7000pm and 5nm becomes 5000pm.
The image is from IBM who made this image of 7000pm transistors to show how small they are now able to make them.
Intel has made some prototypes using an extension of the trigate called gate all around using nanowires. This creates 2 or 4 gates which mitigate the problems given trigate cannot be shrunk past 14000pm easily. Recall the problems Intel has had with the move to 10nm.
The gate all around approach may solve the problem but then moving this from the lab to the production floor is a big step. FinFET has lasted since Intel used it with Ivy Bridge onward.
Some Korean researchers tried to use FinFET at 3000pm but others are working at 2500pm and even 1500pm. Even risk production is a long way off. It’s hard to predict what will happen beyond 3000pm.
It’s hard to tell whether Moore’s law is at the end of the road. No question that the pace of change has slowed considerably due to the staggering cost.