One standard example I use is finding duplicates in a big collection of ite s either via a hashtable or via merge sort. The collection would be big enough not to fit into RAM, so you get swapping onto a hard drive.

Merge sort will mostly just work, because all its memory access patterns are sequential. Hashtables would need to pay for seeks all the time.

So O(n log n) mergesort would be faster than O(n) hashtable approach in this case.

(There are ways to improve the absolute running time further compared to these naive approaches.)