Weird stability issues

My system works stable and runs stability tests like the built in Everest stability test, OCCT or Furmark without a problem.

And then, out the blue, it crashes. BSOD and reboot, both in WinXP and Win7. There's no pattern to this, sometimes it happens after 30 minutes, sometimes after 30 hours.

After a crash like that neiter Windows version manages to boot. They just crash again at various stages of the boot process. At the same time, the memory test that comes with Win7 reports memory errors as soon as it starts. (at 1%)

Turning the computer off and back on - without any time to cool down - immediatly solves the problem.

I still tried to solve the problem with additional cooling, especially of the RAMs, but that didn't help at all.

Before such a random crash occurs, there are no errors in the Win7 memtest. I tried Memtest86+ today and it found no errors either.

The only strange thing I noticed during these memory tests is that the Win7 memtest doesn't seem to run at all with the CPU cache disabled. It just stays at 0% but still reacts to input to change settings or stop it.

I'm not sure if this is a memory problem at all but this still seemed the most apropriate area to post.

Any ideas what might be the problem here are most welcome.

A little update to my problem:

The computer just crashed again and I started Memtest86+ right away.

After 8% it had found over 3000 errors already.

I found it interesting, that the 64bit patterns tested ALL only had errors in the upper 16 bit. There wasn't a single error in the lower 48bits. And the errors where all in certain areas of RAM like a couple of 100 bytes around 250kb, 1GB, 3GB, 5GB, 7GB...

I turned the computer off and back on, startet Memtest86+ again and the errors where gone.

I stopped the test at 12% as a full test takes over an hour and produced not a single error yesterday.

Any help with this would be great. Thanks.

It looks like I found a solution to my weird RAM problem:

I tried a lot things, including additional cooling and all kinds of timing adjustments in the BIOS.

Yesterday I changed the RAM Configuration Option "AI Clock Skew" to Manual and then tried some combinations of Delay/Advance Values for Channel/Bank A and B.

The last try seems to have done it. The computer is running stable for over 15 hours now and passed 4 runs of Memtest86+ without any errors.

Before I changed that option the last few days the computer was stable for 6 to 8 hours max before the problems started. As mentioned before, turning it off and back on solved the problems but only for an hour or two before the memory errors occured again. The longer it was off, the longer the stable period lasted. That suggests some kind of heat issue but I have no idea where exactly. No additional cooling I tried had any effect.

And during those first few hours every day stability testing never forced these problems to appear, no matter how hot everything got. Very weird.

I also have no idea why these RAM issues suddenly started at all, because everything work fine for about two months with Auto Timings before the problems started.

Anyway, right now the two pairs of TWIN2X4096-8500C5D, Rev. 2.1, once again seem to run stable at 1066MHz with 5-5-5-15 Timings at 2.1 Volts.

I just hope it stays this way now. :)

Either way, something like this is really a hard problem to fix because after any change to the configuration, the computer usually worked problem free for an hour or more before the errors started again. And testing the RAM modules seperatley never produced any error.

If you're interested to learn more about these Clock Skew values I found this interesting Article.

http://www.overclock.net/faqs/119265-how-clock-skew-can-save-your.html

I wish I had found this before I stumbled upon this solution myself after painfully long trial and error tests.

But maybe it can at least help other prople with similar problems.

Well, looks like I fas a little too fast with my previous post. After 18+ hours, Memtest86+ once again found some memory errors.

I changed the clock skew values a little more. Let's see how it goes today.

What still puzzles me is the question if this is some kind of heat problem. What part in the Computer heats up so slowly that it could cause such a behaviour, causing errors after several hours of running problem free, even under heavy load?!?

I've encountered lots of computer problems before but never something like this.

A little update:

Today one of the sticks failed completely. Tested sperately it produces thousands of errors after only a few % of Memtest86+ and neither powering down nor changing any BIOS settings has any effect.

The other three modules still test ok, both seperatly and in any combination.

The computer is only running for about an hour since the memory tests so I'll have to wait and see if the defective module caused the other problem too or if the problem is still there.