It’s always marketing bullshit no matter who claims it. TSMC, Intel, Samsung, GF – it doesn’t matter, the contract names haven’t had anything to do with any physical aspect for years. Tweakers wrote here 5 years ago article about, but it lasted longer than that. For a while it was used to refer to at least the smallest ‘feature’, but that hasn’t been the case since long ago.
Intel’s 10nm (P1274) has significantly higher fins than both TSMC’s 7nm and 10nm. They’re 1 nm thicker and are exactly between 10 nm and 7 nm (TSMC’s 10 and 7 are 36 and 30 nm; Intel 10 comes in at 34 nm). After Intel mmm (the distance between two contact points) is only 36 nanometers smaller compared to 40 nanometers on TSMC’s N7 (44 nanometers). and theirs CPP It is also lower than TSMC’s N7. Finally, I’ll add the density; Intel’s 10nm has a density about 25% higher than TSMC’s 10nm; However, TSMC’s N7 is only about 13-14% closer than Intel’s 10.
Tell me, which one is a “real” 7nm? 
No feature even comes close to 7 nm in any process. All this marketing nonsense you shouldn’t pay much attention to. What a knot can do is much less important than the slice baked on it. Nvidia’s Fermi was less intense than AMD’s Cypress, and it didn’t even get close to the same clock speeds, but it still needed a lot of juice. The end result was that AMDs were more power efficient and outperformed (with similar chip sizes) due to their higher clock speeds (GTX 460 vs. 5850; the latter was on average about 10% faster and still about 10-15%). More efficient). Both are at TSMC’s 40nm, but Nvidia has made some mistakes.
So it doesn’t matter what the FAB calls their operations, it’s about the characteristics that their clients can achieve with them. The theoretical properties of large chips such as GPUs or desktop CPUs will never materialize, only small SoCs will ever come close (Apple’s A12 had 75% higher density than the current generation of GPUs). 
†
