UPDATE:
finally got my hands on a multimeter to test the psu. unfortunately, it's analog and it seems like it's worthless for testing the 12v rail.
despite showing massive and frequent fluctuations in bios and mbm5, an analog multimeter shows that the 3.3v and 5 volt rails are stable as hell, as is the 12v rail. however, the 3.3v rail is undervolted at 2.9-3 volts and -tentatively- the 12v rail registers an undervolt at around 10.5 or 11v (hard to tell with analog multimeter)
the reason why i say tentatively is that the settings on this multimeter jump from 10v dc directly to 50v dc and unfortunately the 50v dc setting doesn't seem that accurate; i know that at the 10v dc setting, my 5 volt rail measures exactly at 5 volts, yet it displays as exactly 3v on the 50v dc setting -likewise, the 3.3v rail displays as 1v. in other words, that 11v measurement i wrote earlier seems pretty much worthless.
on top of everything, about 2 hours ago, one molex from this psu that i've been using since day 1 has died (checked with multimeter, no response). i think i'm going to file for an rma unless someone tells me otherwise.
That analogue multimeter isn't sensitive enough; it looks like it isn't properly showing the 12V rail on the higher setting, and might not be accurately representing the 5V and 3.3V rails either.
I don't know how far you can trust either of your voltage monitering tools. Analogue multimeters are somewhat subjective compared to a digital meter, and innaccurate on a large voltage spread. Further, the NF7-S has poor voltage monitering, if I recall correctly.
That said. Those software readouts are a strong indication of a broken PSU. Of even greater concern is the extreme 3.3V rail droop which your multimeter reports... such a signifigant droop on the 3.3V rail can really mess up a system.
Even though the readouts are suspect, I think it is safe to infer that the PSU is malfunctional.
The dead molex is further evidence of a bonked PSU!
some more curiosities:
i cannot run this ram at anything other than cl2.5; even with a severe underclock and extra volts -if i can even manage to save the new, 'looser' timings- i get memtest errors out the *** at about 60% completion. weird. i pulled out the pc3200 valueram (cl2.5 by the way) and am now testing solely with the pc3000 hyperx @ pc3200 with 2.7v, 2.5-3-3-11
Your CAS limitation is on account of mixed RAM; different memory ICs don't generally play well together, especially on the old nForce2 boards, where the board chipset is playing at middleman. Mixing RAM is a love-killer for AXP memory overclocking.
DDR1 is cheap to buy used. If you want to have some fun, you could invest ~$40 in a half gig set of BH-5, and see substantial gains running 2-2-2 with a high FSB on your nForce2 system.
I bet you your HyperX can run 2-3-3-X with some voltage....
just for clarity, how do the specs for my borken psu look (rock solid stable on mm, some variance in bios and mbm5)? also, can someone explain the significance of having one dead power cable (i don't even understand how that happens... as far as i can tell, it's only the one molex that is dead.)? the dead molex coincides with a now almost dead hdd (spins up, spins down, can't get past the ide detection screen in bios) which exhibited zero problems before reaching the current state of inoperability.
Your dead molex is indicative of a dying PSU. Something has blown inside of the unit, which has cut the "juice" ("scare quotes" used because it could be anything that has popped. A capacitor, some circuitry, or similar...) to that set of plugs. To clarify, is it a single plug in a string of plugs which no longer works, or is it an entire string of plugs? If it is only one plug in a string, it's likely a loose pin or cable, rather than some issue internal to the PSU.
Regardless, I'd stop using that PSU as soon as possible, it isn't safe.
The steady deterioration of the PSU's functionality is a clear indication of its being unreliable. It will only get worse, and you really don't want it to take out any of your other hardware with it.
As for the Multimeter reading steady, while your software shows fluctuation and droop... analogue multimeters aren't ideal for reading fluctuation, and it is entirely possible that it is giving you a rough readout, too general for observeable fluctuation.
As I mentioned, I don't recall entirely, but I am fairly sure that the NF7-S did have issues with software voltage readouts.
For posterity, I'd reccomend buying a digital multimeter; they're $10 and wonderful toys