K8NGM2-FID CPU core voltage Fluctuations and instabilities-- Bad caps!!?

Question

I Have K8NGM2-FID bought 6 months ago and started having stability issues since last month.
Here is my configuration
K8NGM2-FID ver 2.0 Bios 3.73 now, used to be 3.5, experimented both, but did not help
Onboard video, Onboard Lan, firewire
Memory : 512 x 2 Corsair DDR400 in dual channel mode at 2.65v 64MB to Video
CPU: AMD64 X2 3800+
HDD WD 160GB SATA150
Winfast TV2000 Expert Tv Tuner
Drivers, tried many version ( currently using latest Nvidia Video, MB drivers)
Now, the description of the problem.
System suddenly hangs with or without load, intermittenly. Nothing to do what we are working on. Even when System is idle.
Hangs means, it could happend two ways,
1. simply it would not respond, screen simply freezes,
2. sometimes, you simply see horizantal lines on monitor( Typically happens when cpu core voltage drops below it can handle). I know this because
I am using RMClock on my Turion laptop, and when we try to find lowest stable voltage we see this behaviour.
Now, some other observation,  from the day one CPU core voltage has been fluctuating a lot. Used to fluctuate anywhere between 1.37v to 1.304v( at default 1.35v).  But still has been stable. Ran 24hour prime95, so no problems.
Recently I am seeing voltage fluctuations is more, as low as 1.28v to 1.39v at default 1.35v, so I tried RMCLock and raised the voltage to 1.4v, but it seems that did not help much.
Now, As per the MSI PCALert4  voltage is changing anywhere from 1.28v to 1.38v, but also I get core voltage field empty( that is right no CPU core voltage shown) and alarm sounds. It seems, CPU does not get voltage at all and gets very low voltage for a second or so.
Things I tried,  Changed the powersupply, changed the Memory, but no use. I am lucky enough to have exact same configuration I built for a friend. So, I checked voltages on his PC and I do not see this behaviour.
Now, I ran out of options.... any idea anyone had similar problem, and if they solved?
Now, here are PCAlert 4 images
http://img171.imageshack.us/my.php?image=128vatdef135sp5.jpg
http://img256.imageshack.us/my.php?image=novoltageif5.jpg
http://img406.imageshack.us/my.php?image=132vredbi3.jpg
Also, another weird behaviour with Bios, even when I change setting in bios, they did not take effect, even loading default does not help, so I have to press CMOS reset button.
Any help will be appreciated.
Thanks

Answer 1

Thanks for the quick reply.
First to clarify, these things has been developed a monrh ago only. It has been fine for more than 5 months.
1). Yes Memory is insalled in closest DIMM sockets to CPU,
 2). I think you mean mem voltage. The memory I have is Corsair Value series and it even runs fine DDR400 at 2.55v, but I bumped it to 2.65v and also tried 2.75v ( 2.65v is typical I have been using with this MB).
3). Memory has been tested with Memtest for 24 hours and ran Dual Prime95 for 24 hours when system is assembled.
Also swapped from another PC I built for my friend with same components( same corsair value series memory) Also tested my memory on his PC, no probs. BTW I also tested my memory at 333, freezing even happens even at that speed.
4). PSU has nothing to do, because I have been monitoring all PSU rails with SPeedFan none of of the rail voltages moved at all. I even monitored with Digital Multimeter to ensure. Anyways I even tried 3 psus, one ANtec 350W smart power supply, ThemalTake 420W, Ultra 400 Watts. All have more than enough juice to power this setup. (onboard Graphics card).
5). Yes, freezing happens in simply desktop too. Nothing to do with CPU load or graphics load. It could happen even when just idling.
6). I use stock CPU cooler in Antec 1650B case with 120MM Fan running in the back( approx 50cfm) and CPU temp never exceeded 47c and MB temp typically around 28c and NB temp around 50-60c using a digital probe and 55-70c using Nvidia Monitor based on load vs no load.
7). As per Dual core patches, I applied AMD Dual Core  optimizer quite a while ago. As per the MS Patch, No I have not applied( But I doubt it helps because it has been stable without that for all these months). I may try that.
 Now, what I am trying, I  disabled CNQ, and logging Core voltage fluctuation with speedfan( no monitoring).
as of yesterday night the core voltage was varying from 1.28v to 1.39v( I think that is insane 110mv fluctuation in CPU core voltage!!?) at default CPU voltage of 1.35v. BTW it has nothing to with CPU load, all these logs generated when it was idle. I remember fluctuation even when MB was new, but lowest I observed was 1.304v.
SOmehow, I suspect VRM is pretty bad on this MB. 
Anyways, see the response and let me know if you think I am missing somethings. Appreciate your help.