pvtu : number of pieces and relationship to parallel visualisation in Paraview

I am learning about vtu/pvtu

I am testing out deal.ii which save data as vtu/pvtu, only one piece is being written out per snapshot.

I plan to test out visualizing the data set on a 4 node cluster (each with GPU) and was wondering if I should invest the time to investigate writing out 4 piece for each *.pvtu to benefit my cluster setup. Does the number of pieces have any performance improvement value if it matches my node count?

Cheers