The M2 Max is closer to the 3060 than it is to the 3060 Ti, let alone 3070. And those numbers are quite possibly overly optimistic; workloads essentially never reach peak tflop, and I would not be surprised if practical workloads are better matched to nvidia's architecture than apple's, if only through sheer industry momentum (But that could go either way).

While the perf/watt is impressive, apple is also using 4 times the number of transistors on TSMC's latest process - and the comparison here is samsungs 8nm, I believe. It's not really all that impressive that that huge silicon investment has some results...

It's a tantalizing hint at what might be possible, but as it stands, I'm not really all that impressed, personally.