[Paraview] nVidia Tesla crash

[Nathan Jarus]

Hey,

I've been testing a render cluster with 4 nodes running 2x nVidia
Tesla C2050 cards. I run an x server on each of these machines, and
have pvserver rendering against each card. I've modified the basic
benchmark paraview comes with to actually stress them (the basic
benchmark tops out at 10M polys/sec at ~30% gpu usage). Everything
runs fine, except...

When I get about 30 max resolution Superquadrics on a scene rendering,
each machine, one by one, locks up hard with just an underscore on the
local terminals. I'm told by people who've done GPU programming before
that this is a GPU hang. That's odd, because Paraview should run
fine...but it's not.

How do I start debugging this?

--Nathan