[Paraview] Using MPI on a single machine

[Mason, Michael (IS)]

Can anyone speculate on what sort of performance increase would be possible if I ran an MPI-enabled ParaView on a single standalone workstation?  The machine is a new iMac with 32GB RAM, 8 cores, and an AMD Radeon R9 M295X (4 GB VRAM).  I'm volume rendering a 4D NetCDF dataset (72 240 MB files).  I'm worried that contention for the GPU among the multiple MPI processes will result in similar (or worse) performance.  Using a single ParaView process, it takes about 15-20 seconds to volume render a single time step and I'm looking for ways to reduce the rendering time.  I'd just like to know if there is even a small chance that I'll get better performance before going through the process of building ParaView from source on OSX.  I've done it before on Linux, so this will be a new experience.

One more question ... Would it be more efficient to combine/concat these individual files into a single, larger NetCDF file?  Or does ParaView perform better with smaller files?

Thanks,
-Mike