There is actually no difference in the memory access pattern of the two approaches, unless your compiler is entirely out to lunch. Memory is only touched in either approach to:

a. loading the address of the next node. b. updating the link address when a node is deleted.

which are performed the same number of times on the same underlying data in either approach.