Solved BSOD Random but Linked to NTOSKRNL Failure

[Arc]

Generally you are supposed to set the RAM to Auto.

If the "memory corruption" bounces back even with only two RAM sticks plugged in, we will concentrate on the voltages.

 

[traderyoda]

Greetings Arc,

I removed two sticks of RAM and ran 4GB in slots 1 (channel A) and 3 (channel B) pushing the machine hard with lots of large memory transfers - no BSODs. I ran memtest86+ through 10 passes - all clear. I switched the sticks I was testing to the other 2 sticks and repeated - same outcome. Then I tried two sticks in slots 2 (channel A) and 4 (channel B) are repeated - this too produced no BSODs and no memory errors.

I then went back and re-installed all four sticks. Since I've come across several folks complaining about memory problems with all four slots occupied I thought it would be wise to go into BIOS and try manually setting values and upping the voltage just a bit. So I turned off auto, set the latency to 5-5-5-18, and bumped the voltage from 1.8v to 1.9v. I also set the freq. to 667MHz since these are 800 MHz chips. The machine booted fine and has been running BSOD free so far. But I have not had the time to really push things yet.

I'll set up the machine for some large memory transfers and run Prime95 and see what happens. But I suspect we're finally closing on the source of the problem thanks to your help!

Cheers
TY

 

[Arc]

Really good news!

Let us know for any further issue.

 

[traderyoda]

Well that didn't take long. I had a lot of stuff open and was viewing some large PDF files in a browser window and POW - BSOD 0x0A again (attached). BIOS is set manually to 5-5-5-18 @ 1.9v (1.8v is spec), and 667MHz.

The machine is stable for routine stuff, but when I push it with all four slots occupied - boom!

Can I back off on latency? Increase voltage?

I really appreciate your patient support - this one is a tough nut to crack!

 

[Arc]

Some motherboards cannot deal with all the RAM slots occupied. Basically it is a fault from the part of the motherboard, but when you can use single channel of RAM, the work will be done.

 

[traderyoda]

Hi Arc,

I think after a lot of work this all comes down to the need to tweak the board to get everything to work. This is an older board and I should have realized that my problems started when I chose to fully populate the four DIMM slots. You pointed me in the right direction and I've now learned that many people have had the same BSOD problems with this exact board and same RAM modules. Evidently the Auto settings in BIOS do not deliver a stable state if all 4 slots are filled.

Here's what I've done. I did confirm this board will work with all slots populated - but tweaking is required. I've manually set the RAM voltage and timing (same standard timings in the Corsair specs), boosted the voltage up to 1.9V (which is actually 1.98v since this board overvoltages by .8v). I set the freq. to 667mHz and not the 800MHz the RAM is capable of. I also boosted the Northbridge voltage to 1.34v which is plenty safe. Evidently the P5Q series of boards need a higher Northbridge voltage to adequately support 4 chips. Finally I set the AI Clock Twister variable to "LIGHTER" to see if that helped stability, and I set a skew of 50ps in Channel A with Channel B left at normal (evidently timing unison between the two channels is one of the leading causes of BSODs). None of these changes caused any noticeable changes in operating temperatures.

I have all the RAM in now let memtest86+ run for 10 full iterations (all tests) and ran Prime95 without any BSODs. I may have to tweak some more over time, but it's clear that these tweaks are necessary to get a stable system.

THANKS ARC FOR ALL OF YOUR HELP!!! I now have a stable system and have learned a ton along the way.

Cheers
TY