[Paraview] CAVE + ParaView volume data visualization

[Emiliano Pastorelli]

Hi,

i'm trying to use paraview to visualize a nhdr+nrrd file representing 
the CT of a cylinder of concrete containing metal fibers. The file is 
definitely quite big, more or less 800x800x600, and what i need to do is 
to find a threshold in the volume to highlight the fibers. So far so 
good, i works, but defining it slow is not enough.

I thought my linux machine should be powerful enough, i have two 8-core 
cpu, 4 nvidia quadro 4000, and 64 Gb of ram. When i run it on a single 
machine (with mpirun -np 16), it's really slow in showing the volume, 
thresholding and once thresholded in moving the fibers around.
But when i try to run it on two display, in a temporary CAVE test-setup, 
it becomes so much slower, that to try to track the user on it becomes 
almost useless, as every movement takes a few seconds to be made.

Now, what i noticed is that the CPU are shooting up to 100% as soon as i 
deal with this datasets, but the graphic cards are almost unused (check 
with nvidia-smi). The top i reached was one of them using 160Mb of RAM 
(over 2 Gb available). Is that normal? I checked the ccmake i used and i 
didn't even activate the mesa libraries, so paraview should be using the 
nvidia drivers.

Any idea on what might be causing this slow-down? I mean, the dataset is 
big, true, but i thought this machine was strong enough to support it!

Best,
Emiliano