For example if we look at the 22nm processes, TSMC, Global Foundries, Samsung and a few others stuck with MOSFET for 22nm. While there was an increase in density that it was on par roughly with Intel’s 22nm, Intel had switched over to FinFET for their 22nm. The MOSFET 22nm performed so poorly that even with the increased density, its performance was almost as good as the 28nm process these companies had.
And those companies then switched their 22nm processes to use FinFET, they didn’t have any increase in density for their processes, but since it performed way better than their 22nm, they decided to rename their 22nm using FinFET to 16 / 14nm. That is where the disconnect between Intel’s 14nm / 10nm and TSMC / Samsung / other’s 14nm / 10nm / 7nm came from.
Tho it looks like that Samsung’s 3nm from what I can find is going to be roughly the same density as Intel 4 and TSMC’s 5N. But Samsung’s 3nm will be using GAAFET, which likely will have it performing better than those nodes, even at the same density.