The simplest way I can think of to put it is thusly: Cache memory hits most of the time If the L1 cache misses then there is an equal chance of the Lc cache having what you need. My next point is best illustrated with math. Let us assume a 90% "hit ratio" 90% + (90% of 10%) = 99% Therefore, 99% of the time, the data you need is in your cache so your CPU doesn't need to call on your (comparatively slower) RAM. That is better than the 90% that you would have with no L2 cache. However, if the L1 cache misses, there are FAR more possibilities, MANY more guesses, so the L2 cache MUST be bigger in order to holed the larger number of guesses.