I think one example for pre-fetching benefit for shared cache is data-parallel model. Threads running on different cores could easily pre-fetch data into cache that would be used by other threads.

are there examples of this type of processors that aren't gpu's?