Frequent, but inconclusive, BSODs

[MagicAndre1981]

Could you elaborate a little why Aero rules out CPU?

because Aero Glass is rendered by the GPU and not by the CPU.

[dencorso]

dencorso I don't have those keys at all, will setting them to 0s and then deleting (if it doesn't help) return them to their default values?

Yes, assuming that Windows 7 x64 still honors them as older versions did. That's why I call it a longshot. But since its easy to do and reversible, I think it's worth trying.

Even if you don't see any effect immediately afterwards, do wait a couple of days before reversing, just to be sure.

[wlw]

Tripredacus - no, I haven't. Previously I had to disable USB3 controller because the driver was crashing, now I'm also disabling core unlocking, HPET, C1E and on top of that downclocking and overvolting... so I have to disable more and more for it to work, which is nowhere near how a new PC should behave. The features are there to be used and I have to disable half of them to be able to actually use the computer, this is p***ing me off. Disabling another feature is not the way to go, not unless I know what is causing these problems and how to fix it.

That being said, since I have disabled C1E and HPET and given NB a little more juice, everything is working fine. That makes 3 days of solid stability in every scenario that used to crash the PC - idling, streaming videos, gaming etc.

However because it has the tendency to go quiet for a couple of days and then crazy again, I'm withholding my optimism and refraining from doing any changes until the above tweaks prove to be the solution, at which point I'll try to narrow it down to just one factor, probably undervolted North Bridge.

If it happens to be just another passing peace, I'm going to throw everything at it, including Aero

In the meantime I'm still looking into getting another CPU and GPU.

[wlw]

So, ahem... I switched the GPUs from my 6850 to 5670 and reset the BIOS to the bad* settings and everything is running fine.

*) C1E [Enabled], HPET [Enabled], NB VID 1.1V.

I'll give it some more time of alternating stress and idling to confirm, but still... It doesn't really make sense if my GPU is faulty. There have been indications of low level hardware problem, yes, but besides the occasional strange pattern on the screen, nothing really points to the GPU. Even more, I can stress it all day long and overclock it to it's limits and it's still working fine and then bam crash when idling on 100/150 clocks. The CPU is much more likely to be the cause but unless there is something very complicated going on, does anyone see any particular indication as to how the GPU could cause all those above problems?

Deja vu, double fault again on amdppm.sys:

X64_0x7f_8_amdppm!C1Halt+2

Disabled C1E yet again, this somehow drives the suspicions away from GPU.

I've been looking for this driver related problems and found Andre talking about it on another forum (here) so why not, I'll go all over the RAM again because I'm out of ideas for now.

Update:

I've been testing the RAM sticks one at a time and both of them rarely give errors. 99% of the errors appear on test #5 - Block move. Youngest bit of the oldest byte only, so it's 0xEFFFFFFF instead of 0xFFFFFFFF. Once there was an error on test #7 - Random pattern, and it was the same - oldest byte was off by 1 - 73e9ab6d when it should be 63e9ab6d. So it's one bit in the oldest byte, which is negated sometimes (7 = 0b0111 instead of 6 = 0b0110 and E = 0b1110 instead of F = 0b1111).

There can be numerous passes without errors, but when they do pop up, it's in the first pass and they span the entire stick. This makes me wonder if this is in fact a problem with the memory or perhaps with the memory controller in the CPU?

Edited June 9, 2011 by wlw