All of these have been around for a long time, but just like well-maintained software, 7 years worth of incremental improvements add up:

Since RAM (slow) and/or cache access is involved in nearly every step--which becomes increasingly complicated when trying to preserve cache coherency across multiple cores--improvements in the next two are a big deal:

https://en.wikipedia.org/wiki/Memory_management_unit

https://en.wikipedia.org/wiki/Cache_hierarchy

Executive summary: Work smarter, not harder, & the ultimate measure of performance, is performance (which may sound stupid, but it's in the textbook, because it's true!).

I would also add that more cores aren't necessarily better. The utility depends upon the nature of the task & how memory-hungry it is. If the task is inescapably sequential, it doesn't really matter how many cores are on the die. Same story with parallelizable tasks that pound RAM: at any given time, one core is hogging the memory bus, and the rest are waiting their turn. They may take turns, but at any point in time, it's essentially single-core performance.

The place where multi-core really shines is when you have a highly parallelizable task, where each thread grinds really hard over a smallish data set that fits comfortably in the core's cache. In that case, you can definitely max out all cores. Though from what I see in the wild, that is a rare case.

A lot of the industry is really just gaming benchmarks at this point, which are, for the most part, bullshit. I think Apple will remain in very good shape on this front, if only because of their customers (i.e. normies instead of gamers; people who haven't fallen completely into the quantity cult). They will complain when it stops feeling fast (the only measure...), instead of taking it as a challenge & wasting their lives on overclocked water-cooling bullshit.