More On Video Card

After several more hours of testing and researching, I had a few more observations:

• Skyrim utilizes DirectX 9 and possibly some DirectX 10.
• Grid utilizes DirectX9 and possibly some DirectX 10.
• Battlefield 3 utilizes DirectX 11
• Batman: Arkham City utilizes DirectX 11

So far I've only experienced problems with Skyrim and Grid (the latter only in the past two days).  The crashes are the same: video feed goes out, sound loops, system unrecoverable.  This can happen after several hours or after just a minute (or even straight from power off, boot, game load, crash).  The usual suspects are out:

• Heat (measured over and over, max GPU is like 72C and max CPU is 68C.  This was extreme testing.  Normal operations is about 35C CPU and 42C GPU lower.
• Overclocking - the crashes are seen without overclocking, with overclocking and with underclocking.  It ain't the overclocking, okay?  I tested it.
• Malware - no malware on the system.
• System errors - NO ERRORS IN THE OS EVENT LOGS.
• Power Supply - replaced 750W power supply with a Silverstone 1000W power supply

Now, the fact that no issues are seen in Battlefield 3 and Batman and only the games NOT using DX11 led me to this article which describes how installing an older DirectX 9 game can muck up your DirectX 10 but leave DirectX 11 intact.  Weird, that sounds like my issue!

Quote from article:

Obviously, something is wrong with the DirectX 9 redistributable package. However, I don't think Microsoft can do anything about this since the DirectX 9 redistributable is everywhere and has been in use for ages. We just want to warn you about this problem so you won't need to scratch your head if your DirectX 10 game does not work. It could be because you just installed a DirectX 9 game. Try reinstalling the DirectX 10 redistributable.

Well, I tried the fix (install DirectX 10 runtime manually), but to no avail.  I still get the crashes in Skyrim and (now) Grid.  YARH (Yet Another Red Herring)?  Likely.

Frustrated and unable to get away from the observations:

* Crashes in DX9/10 games (Skyrim & Grid)

* Does not crash (yet) in DX11 games

* Just started crashing in Grid a few days ago

* No errors found using OCCT, Furmark, Heaven, etc, etc, etc

* No OS errors (this means the OS does not even catch the error, indicating it is likely NOT the video driver).

* Was able to change the crash slightly by disabling Timeout Detection Recovery (TDR).

All of the above points to a video card that is not behaving properly. 

Frustrated, I removed the Gigabyte GTX 570 from my system for packing up for the RMA and I found an older Nvidia Quadro FX 4600 sitting around, so I installed that.  I tried and tried to replicate the crashes, but I could not.  Changing the video card eliminated the observed crashes.  The logic is inescapable: it's not software, it's not my setup, it's not my configuration.  The only variable that changed was the video card itself.  Observation: video card is causing the problem. 

Guilty? We shall see.

It's weird that this would indicate that so many people have faulty video cards.  But then again, is it really so many?  In hindsight, no.  Sure, there are a lot of forum posts about it, but then again there are probably a lot of forum posts about....well, anything.  More than half of the posts tend to be trollers, leaving only a few with the same actual issues.

And then I remembered how my video card got a low rating (around 68%) for the ASIC in GPU-Z.  Perhaps it's true: Skyrim does push your hardware in ways other games and stress tests do not.  Here's another interesting fact: Had I not played Skyrim, I would have had zero issues with this system.  Well, at least I think so.

This is a very good video card.  When it works.

Hopefully this is the final test.  It is important to note that prior to the video card swap, I was already up to over 10 crashes in one day.  It was getting crazy.  After the swap and after several hours in Skyrim, loading saved games that would consistently crash, I got zero crashes with the older card.  

The GTX 570 now sits in a shipping box, ready for RMA.  However long it takes to get there and for CyberpowerPC to send me a replacement will hopefully be worth it if I've found the root cause of the problems for the last four months.  Now I just hope CyberpowerPC doesn't get the card, run Furmark against it and then send it back to me.  If that happens, I may just have to get another video card altogether.

"Video card, huh?  You were losing the video feed?  And then crashing?  Must be tough to be so frigging smart!"