The Intel Prescott is one of the hottest processors money can buy despite the die-shrink. There is a condition called "Threshold Voltage" which is the amount of power necessary to turn on and switch the transistors gates. Prescott switches their gates much quicker than their predecessor "Northwood" and Intel did not take into account the leakage of the switching gates power or the wire delay for which Prescott is infamous.
There is a formula for switching power:
Power ~ ½ CV ² Af
This formula states that dissipated power is linear with the effective capacitance, activity and frequency. Power increases quadratically with the CPU's core voltage. Thus when you overclock you are increasing the necessary voltage by a multiple of four, not one for one and with the already heady requirements of Prescott it is necessary for a strongly functioning power to supply these needs.
Now leakage power is a fundamental sieve with relation to dynamic power because the leakage power increases exponentially not linearly. Remember when you watered the lawn and there were tiny holes in the water hose? How they somehow got bigger as time went on? Of course the bigger the hole the more water you lose. This is how leakage power works and is part of a current so one (at this time of technology) can not fundamentally stop the leakage, one can only hold it in some form of abeyance.
Then added to the power loss (and hence more power of necessity) there is an added effect called "Wire Delay" which is the effect of the switching speeds and the resistance of the capacitors (RC). Now as you pump up the voltage the RC becomes more of a power bottleneck and thus needs more voltage for the signal to travel.
So the Prescott's failure (and it can be called nothing less) was the exploding leakage power with wire delay even though it was somewhat contained by additional technology introducing high-K materials and SOI (Silicon-on-Insulator), however this did nothing to slow down the power hungry requirements of Prescott and served only to allow the substrate to accept such power.
So your overclocking of the Prescott necessitates a very decent power supply and one which can accept both a linear quadratic (^4) raise of peak along with an exponential power leakage dissipation.
Originally Posted by ddedicate
However, can you USE ANY 450W PSU like the NSPIRE model NSP-450P4DL? What is your opinion on this puppy?
I can not say as I have not used that PSU and one can play games with voltage/amperage readouts. They do not take into account spike term fluctuations etc so I will only advise on PSU's I have used and tested. You might wish to post in the Power Supply section to determine if this is indeed a decently functioning PSU. It has 40A on the +5, 20A on the +12 and 28A on the +3.3 rail. Keep in mind that the +12 is a single rail. For me the +12 is a bit light but I have never used it so can not speak with any real knowlege of this PSU. The Fortron has dual +12 Rails with 18A on each 12+ rail.