Easy
Typical access latencies from L1 through DRAM
A pointer is chased through progressively larger, uncached working sets so each access is forced to hit in L1, then L2, then L3, then main memory (DRAM) respectively. Which ordering of typical latencies (in core clock cycles, roughly, on a contemporary desktop/server core) is closest to reality?
8 of 47 in Computer Architecture