PVTU or HDF5 for efficient solution visualization


I was reading the VTK documentation and I came across a couple of VTK formats for parallel computing. One such format is the pvtu and the other is HDF5. Which is the better or efficient format for large data visualization? The VTK file format documentation suggests the pvtu to be a better format but I’d like to hear from the community/developers.

Also, I’ve come across blog articles which describe streaming based visualization (from @berkgevec I think) for efficient visualization of large datasets. Can this be done for the two formats?