I'll simplify it. RAM is somewhat random access. It is better than Flash memory with Flash's multi-KB minimum erase size. But RAM has a huge delay for the 1st 8 byte random read/write. The next 4 to 8 (depending on RAM design), 8 byte blocks, are free. If after the 1st 8 byte block, you request another block, before you get 1st block of 2nd random request, the time to have delivered blocks 2-4 of the 1st random read will have gone by with the RAM bus being idle. You might as well have listened for blocks 2 thru 4 and put them in CPU cache speculatively, otherwise you just wasted RAM bus time.

Core 2 CPUs have a cacheline of 64 bytes. That design hints that the CPU will read sequentially all the data that it can (16/32/64 bytes) so the RAM bus doesn't go idle. It also means to read 1 random byte from RAM is to read atleast 16 bytes from ram. You might as well use bytes 2 to 16 for something as a C programmer.