Following are the specifics of the computing environment I am using:
System configuration : Foxconn Destroyer Motherboard with AMD Phenom 9850 Quad, 16GB RAM,
4 Tesla c1060 attached to the PCIe Gen2 slots and nForce 780a onboard for primary display.
Operating System: Windows XP professional x64 with SP1 and SP2.
Host Developement Platform : 64 bit configuration in MS Visual Studio 2008 with SP1 installed.
I have used CUDA 2.2.
I am trying to determine if there exist WatchDog problems with WinXP 64bit +SP1 + SP2.
The code I have written uses OpenMP and each of the 4 Tesla GPU is assigned to a dedicated CPU thread.
I run a cascade of loops in the kernel and record the time taken by the kernel to complete.I allow only
one CPU thread to process at a time which implies at any point in time my process is running only
on one GPU. I donot use nForce for processing as it has the kernExecTimeOut field as 1.
My system gets “shut down” randomly when I run the executable.
I have run the executable several times now and in some runs upto 2 Tesla cards gave
output before system got shut down. But never has the executable run to completion.
I donot know what I causing the system shut down(no warnings,no messages, no Blue Screen.
Simple Power Off ! !)
Attached is the code of the program I have used.
WatchDog_Timer.txt (9.72 KB)