Ep45-UD3P freezes when changing NB volt

Something strange sometime happens with my motherboard when I change NB volts.
I'm using Easy Tune to tweak my settings in windows, and i can change all the voltages fine, except the MCH voltage. If I move up too much, or when really pushing the other voltages, the PC freezes and a shutdown is the only way out. Sometimes it wont even boot after that, and just starts turning itself on and off until I shut it down. I then have to take out my RAMs, pun in another stick of RAM (Another brand/spec) and then it will boot up. Then I can shut down the computer, install my regular rams again, and start all over.

For example, I wanted to really push things so I went from 1.34v til 1.4V MHC and diddnt change anything else. No lock up. I then changed all the other ram settings ( PLL, VTT, Vcore and DRAM). Still no lock up. Then I changed the MHC voltage to 1.58 and it locked up as soon as I hit the "Set" button in ET. And I'm still running at stock frequency btw, only changing volts here.
After a reboot I tried going directly from 1.34V to 1.50V and that gave me a lock up also.

The same thing happens if I change the settings in BIOS. Then there is just no post instead of the lock up. And the computer starts rebooting endlessly.

My setup:
Ep45-UD3P
QX9650
2x1Gb Ballistix DDR2 (D9 chips)
GTX460
Corsair TX750

Tried with W7 64 bit, Win Xp 32 bit and Vista 32it.

Any ideas?

Just curious as to what FSB and DRAM frequency you're pushing in that you need to supply a vMCH as high as 1.5 - 1.58V on a P45 w/ only 2GB of RAM?

yeah wow! I hope that NB has a fan on it with those volts!!!

There is a Ulta Kaze fan blowing 5-6cm above the Nb heatsink, and the temperature are fine.
I'm struggling with 520-525Mhz FSB on my QX, and the rams running 5:6.

But thats not really related to why I made this post Any ideas to why it locks up when changing the volt? It locks instantly, or within 1 sec when doing it in windows, and just wont boot at all if I make the changes in bios.
Could MCH/DRAM reference play a role in this?

Last time I remembered that I did remount that NB + Mosfet heatsink and apply good tim on that nb chip, and resulted better oc ing fsb, but not much though.

What interesting is after the re-tim, the nb heatsink is getting hotter significantly than before at the exact same oc setting, this could be caused by better contact, hence better heat transfer to the sink, worth a try.