N5 is already showing better yields compared to N7 and N10 going by production times, and TSMC also plans to squeeze two more nodes in between N5 and N3. N5P is supposed to start volume production some time in 2021, while N4 should become operational in 2022. The surprising part is that N3 scheduled for 2H 2022 will not use GAAFET transistors. Instead, it will be based on some sort of improved FinFET tech.
https://www.notebookcheck.net/TSMC-releases-more-details-regarding-the-upcoming-3-nm-node.489515.0.html
Wow. More performance, less power, smaller, based on well understood architecture. Doesn't look good for Intel, as each generation the gap will become wider and significantly more difficult to bridge. The best Intel can hope for here is to remain within 1 node of TSMC so consumers don't notice the gap, Intel 7nm verses AMD 5nm etc.
Well TSMC certainly rules chip making, but there is a lot of marketing hype in the 'nm' nomenclature. As marketeers are apt to do, that is 'stretch the definition'. At 20 nm and above everyone was basically using the same terminology, now many are stretching what the 'nm' rating means. You have to do really detailed research to find exactly how to compare various 'nm' claims. That said, TSMC is still leading the push lower no matter how they name it.