Linux system crash/freeze. How to isolate GPU vs. driver vs. linux xserver/kernel

I have the nvidia-debug file to attach, as soon as I can figure out how to do that… The symptom is a sudden “freeze” of the system: the monitor screen goes blank although monitor status still shows DVI Active Signal; there is no keyboard function, even “Caps Lock” does not respond with LED indication; no evident mouse response either although with no monitor display it is impossible to know; unable to connect via LAN port, for Telnet or FTP probing.
So, then, the only recovery is a System Reset and reboot. I suppose that destroys most of the utility in the files and status captured by the nvidia-debug script but there is no other recourse. I hope some data might be retained in the GPU through the Reset since the power is not cycled off.
I have tried two different motherboards, with the same results. The “crash” is not immediate. The application running is a heavy computation OpenCL task (Seti) and typically things run normally for 4 to 20 hours before the fault.
(1) Is there such a thing as a (Linux) diagnostic program for the GPU (GTX 750 Ti)? Something that would push it to the limit and capture any abnormal condition rather than just lock up the system.
(2) If I set up a Telnet connection to another system, is there anything that can be sent, on a periodic basis, to serve as a “post mortem” log of conditions leading up to the crash? The LAN connection will be lost, of course, but maybe the last status message could be useful.
… look for an attached file …

nvidia-bug-report.log.gz (123 KB)