Pat said:
I am having problems with a computer crashing during what could be referred
to as times of "heavy graphics processing".
The self diagnostics that came with the card are coming back OK. The card
seems to work OK for most applications. All the latest upgrades and patches
are installed both for the card and XP.
The problem seems to occur when the computer is performing heavy graphics
processing. It simply locks up. No response of any kind. The video freezes
until you hit the reset button. I am unsure if the problem is in the video
card or the motherboard as I have no way of accessing the computer itself
with a diagnostic after it becomes unresponsive. The dump logs haven't been
too helpful up until this point either.
I'm looking for some general suggestions or hints in narrowing down suspect
components.
You can try running this program as a stress test. The
Prime95 program (provided by Mersenne.org) has a Torture
Test option, which carries out a math calculation with a
known answer. The program can check the results for
correctness, and since the program uses both the
system memory, and the processor, will tell you whether
the core of the system is healthy. (If the program asks
to "join GIMPs", say no and carry out the test. The
default memory allocation should be OK, as it seems
to leave enough room so you can do a few other things
while the test runs.) This version is multithreaded, and
should start a thread per core.
http://majorgeeks.com/Prime95_d4363.html
The reason I'd suggest running that program, is to distinguish
a CPU/memory problem, from a graphics problem. If Prime95 can
run for four hours without stopping with an error, then you're
in pretty good shape. I've had unstable computers, where an
error was detected in under ten seconds. If Prime95 is clean,
then that points more towards the graphics card, graphics
card driver, or the power supply.
I don't really have a good diagnostic for graphics cards.
There are certainly graphics related programs you can run,
but I don't know to what extent they really test the
video card. I use the 3DMark demo loop for simple
stability tests (there are a number of versions of
3DMark, and some are a bigger download than the
others).
http://majorgeeks.com/3Dmark_d99.html
The graphics card can overheat (like if the fan is stuck),
as can the CPU. "Speedfan" from almico.com, can give you
some temperature readouts (even extracting hard drive
temperatures via the SMART interface). As Brett points
out, there are other programs that can read out the
GPU temp. 65C is "hot enough" for a processor.
GPU designs seem to allow more than that.
To determine whether it is software related, I like to
boot an alternate OS and try to test there. While I use
Linux for its "boot from CD" convenience, it doesn't
really have anything that approaches 3DMark for a
graphics test. (GLXgears is an example of something
you can run, but it isn't much of a test. I have run
a copy of Quake Arena in Linux, but for that you need
the data files from the game CD. I used a Macintosh
CD, to get some donor data files. Linux even has
some code to read Mac discs.)
If games fail under both OSes, then you'd know you
have a hardware problem of some sort. It isn't
likely that the exact same software problem would
exist in both environments.
There is a test you can try, when a computer "locks up".
If you have a second computer, you can open a command
window, and try "ping 192.168.1.2" where the number
is the IP address of the stuck computer. If the
ping command gets an answer, that tells you the
CPU on the "frozen" machine, is still running the
network protocol stack, and isn't really frozen.
There are probably other networking functions you
could set up and test as a means of verifying
the function of the frozen computer. Like a
remote desktop session of some sort.
Paul
