As the saying goes, "memory is the new disk". Data structures like B-trees which were designed to improve the performance of data accesses from disk are now useful for improving the performance of algorithms in memory.
I don't have any data readily available, but it's generally understood that you want a cache-aware sort, yes. Usually I see people opt for samplesort rather than sorting chunks and merging, though.
That is, use quicksort for chunks up to the size of the L1/L2/L3 cache (depending on benchmarking) and then mergesort the chunks.
[1] https://en.wikipedia.org/wiki/External_sorting