I've been punishing myself and Oblivion on Win7 -- actually am getting to the point where I can *start* doing quests, for 20-30 minutes at a time before a crash...need to get a new graphics card before I am done (current one hits 99C....have a feeling that's contributing to instability.)

But looking at my Windows event log using the windows application log using 'sysinternals'
psloglist -- and a perl and a bash shell script under cygwin, I detected some distinct patterns
in the crash behavior.

Looking at errors sorted by module-errorcode-offset, I could see some of them look random...but
looking at the just the ones with error counts (this is over the last few weeks), >= 3, I have 29
sources.

Of those, the top 4 account for over 50% and and the top 10 account for >85% of the crashes. So
given the frequency of these type 10, I was wondering if anyone has ever correlated crash addresses with
points in the code that could tell me what to address?

Here's my stats for the top 10 -- columns are "running % of total, running sum of crashes, #crashes for 1 cause (high->low),
and the 4th column is the "module" (if it was a dll, then lists dll, else lists Oblivion.exe), errorcode (mostly the same, and the interesting part -- the offset. I figure with the tops out of 928 crashes, accounting for that might be a good place for
a 'dll' examination. Not sure how possible it would be, but a DLL could sit in memory, and make specific patches at
the most common error points -- it yield features like:

"module A requires B, but B hasn't loaded yet..(crash), or modA is looking for file X and it's nowhere. OR Your char requires formid X, (reference DB of known extensions & formids), this likely means you need to have mod'X' loaded to get that
working again... Etc...

Or better yet... but more difficult -- address the error condition in real-time...

ok... probably not even close to worth the work... but depends on how many people run into these errors and I have no clue if there has ever been any attempts at correlating crash addresses with known problems.

Anyway My top 10:

99C? "Contributing"? You think maybe?

GPU start to fail at the 99/100C mark. My previous GPU (Radeon 4870X2) would start displaying graphical errors at 100C and then reboot the system at 109/110C. Those errors are most likely just

Access Violation Error

Graphics memory can fail before the GPU Core detects an overheating condition. Most likely this is the case.

Just purchase a new GPU. I'm happy with my Radeon 6950 which idles at 38-42C and maxes out at about 60C (so far in games, not a load test).

Also, the longer you continue to let the GPU run at high temperatures, the more likely you are to cook something nearby on your motherboard. Repeated heating and cooling (along with associated expansion and contraction) won't do anything good for plastic insulation, solder joints, etc.

I replaced my old card when it was consistently hitting 90C. My new card (a GTX560) rarely goes above 55-60.

----
This is a GTX 590, and lately, hasn't been getting to 90C as I've been turning the graphics settings down on OB to try to get stability.
But a GTX590 is a 1 card SLI of 2 GTX 580's I believe... it should be more than enough card for a 6 year old video game, no?

Nevertheless -- just have to fill out the RMA and they'll send me a new one -- already contacted them about it.
So "it's in the works"....

It consists of two GPUs that are stripped down from the GPU in the 580s. The Asus MARS II uses two of the GPUs that are not stripped down, consisting of the GPUs from the 580s.

Anyways, dual-GPU cards can be problematic at times so I'd personally just keep a close eye on things just in case.