Isn't fast on-die/on-package (not sure either) memory access a key factor for Apple's gigantic performance leap, though? With memory pre-soldered on small notebooks for years now, it's not much of a difference for consumers anyway once more than 16GB becomes available.
I'm not sure that the soldering does much to improve memory access.
The anand tech M1 article measures memory latency at more than 90ns [1], which is almost twice what I see for AMD and Intel benchmarks, at 50ns and 70ns [2]. If these are not comparable measurements, I'd love a correction!
It seems to me that there's significant room to improve the memory subsystem for Apple Silicon to reach parity with desktop RAM performance.
