I've been digging into the Apple M2's cache architecture and there are some intriguing performance implications—what do you think?