Yes, but that's not at all how we're packing bits into physical chips right now. The fact that access latency is O(N^[1/3]) has nothing to do with this, it's ju

Yes, but that's not at all how we're packing bits into physical chips right now. The fact that access latency is O(N^[1/3]) has nothing to do with this, it's just that the relative size of caches and memories have been designed that way. That this latency is of the same order as if you were putting bits in a cube is more coincidental than anything.