dumping core during isosurface routines

[Mark J. Oliver]

Looking for some pointers here.  I've got a collection
of GL routines that appeared to work fine on a variety of
data until a few days ago.  As my simulation progressed
the isosurface started breaking up and then the machine 
(screen) would blank, flash back on with an indication that
the news server had received a signal 15.  Nothing could
restore the news server.
Looking in /usr/adm/SYSLOG I find:

Mar 10 11:25:07 bambam grcond[4413]: CIO: GE FAULT:vertex buffer: 2 words
Mar 10 11:25:07 bambam grcond[4413]: CIO: 0001 0001 
Mar 10 11:25:07 bambam grcond[4413]: CIO: 
Mar 10 11:25:09 bambam grcond[4413]: CIO: GE FAULT:vertex buffer: 2 words
Mar 10 11:25:09 bambam grcond[4413]: CIO: 0001 0001 
Mar 10 11:25:09 bambam grcond[4413]: CIO: 
Mar 10 11:25:09 bambam grcond[4413]: CIO: GE FAULT:vertex buffer: 2 words
Mar 10 11:25:09 bambam grcond[4413]: CIO: 0001 0001 
Mar 10 11:25:09 bambam grcond[4413]: CIO: 

and also


Mar 12 20:44:55 bambam unix: Graphics timeout (FIFO still > half full)!

So any ideas on where I should start looking?

Thanks in advance for any ideas
Mark Oliver  oli at inel.gov