Need help resolving IC7 stability issue

I'm having an interesting stability problem with my IC7 overclocking. Not sure where the problem lies. Here is a rundown of my setup:

Antec 1080AMG case w/ 480W TruePower PSU
Five 36CFM 80mm case fans (2-rear, 2-front, 1-side panel)
IC7 mobo
OCZ PC4200EL 2x256mb
Asetek Waterchill KT12A-L30
eVGA GeforceFX 5900 Ultra at 516/943 Mhz (7314 3DMark03)
2.8C P4 @ 3.49Ghz
FSB 249Mhz
Idle CPU temp at 26C system temp: 37C
Full Load CPU temp at 26C system temp: 41C (warm day today)
Currently watercooling all three chips (CPU/NB/VGA) in that order.
Idle GPU temp at 26C: 41C
Not sure what NB temp is.

The problem I'm experiencing is about once a day or two, I get a BSOD with a Invalid Process Attach or IRQL type error. If I lower the FSB to 245 the problem seems to disappear. The funny thing is, it will pass the memtest86, Prime95 torture test, and 3DMark03 for long periods of time and never fail, leading me to believe the system is stable. But then after a day or two with it just sitting on the desktop it will crash in the middle of the night. I would like to know if this might be a motherboard stability issue at that high an FSB. Might this be a problem that only a voltage mod to the motherboard would fix. Any insights on this would certainly be appreciated. Thanks.

it could be a vtt, (termination voltage to the ddr), or a voltage issue to the ddr, or maybe u need to bump the agp voltage a tad( the agp voltage also feeds into the northbridge circuitry on the ic7 boards) try bumping up the ddr voltage and agp a little and see if it helps....

I will definitely give that a try. I'm curious though, would the AGP/DDR voltages be a factor here considering this happens sometimes while just sitting on the desktop? By the way, I'm running with the latest Forceware 60.72 drivers, but have noticed it as well with earlier versions.

another thing to consider is nb temps. if ur borderline it could be a factor also

Since I'm watercooling the NB, I don't think that would be a problem. I did get the Invalid Process Attach attempt again around 5:30am this morning. It had been running fine all night. This was after raising both the AGP and DDR voltages, so I guess that's not the problem.

sounds like something ur tests are missing is failing...