VI7 BIOS 16 Not Maintaining CPU Voltage Changes

ENERMAX EG365P-VE-FC (350W) PSU
ABIT VI7 (BIOS 16)
P4 2.4(B)GHz 133MHz (OC @ 166MHz FSB, 1:1 w/ memory)
COOLMASTER Aero 4 HSF
K-BYTE 512MB (2 x 256MB) DDR PC2700 (333MHz)
SEAGATE BARRACUDA IV ST360021A (2 x 60GB) (RAID 1)
WESTERN DIGITAL WD800JBRTL 80GB
PROMISE FASTTRAK 100 TX2 ATA/100 Raid Controller
D-LINK DWL-G520 802.11b/g PCI Wireless Network Adapter
LITE-ON LDW-411S DVD±R/±RW CD-R/RW
LITE-ON LTR-48246S 48x24x48 CD-R/RW
TEAC Floppy
PNY GeForce4 Ti4200 64MB DDR AGP 8x
VIEWSONIC VG-151 15" LCD Monitor
ALTEC LANSING ATP3 Multimedia Speakers
HP DESKJET HP895Cse Printer
WINDOWS XP PRO (SP1)

I've been running the above for about three months. Prime95 has been run
from time to time, and always runs for 24 hours, no problems. I just ran
Prime95 the other day (after not doing so for a few weeks), and now it
consistently fails, and quickly, usually in 10 minutes. I ran memtest-86
several times, the memory is fine, no errors.

So I went into the BIOS PC Health and noticed the CPU voltage is unusually
low (~1.47v, 35C) @ "By CPU" voltage setting. So I increase it to 105%, and
after reboot, it reads ~1.53v (39C), looks good. I rerun Prime95 and it
fails again. I check the BIOS, and low and behold, the CPU voltage is back
to ~1.47v (35C)! I check the CPU voltage setting, it's still 105%.
Curious, I max'd the CPU voltage to 115%, rebooted, and the BIOS PC Health
reports 1.66v (44C), again, looks good. I reboot, check the BIOS PC Health,
and it's back to 1.47v (35C)! And again, the CPU voltage setting is
unchanged, it's still 115%.

What's going on here?

I did a little more experimentation and discovered that if I change ANYTHING
in the BIOS, the CPU voltage in the PC Health reports expected/reasonable
values (i.e., voltage and temps consistent with the CPU voltage setting).
But as soon as I reboot and check the BIOS again (no changes), it returns to
nominal values (i.e., 1.47v (35C)). As a convenience, I kept changing the
CPU warning temp (could have changed anything, results are the same), and as
long as I changed it, each time, the BIOS PC Health would report the
increased CPU voltage (and temps). But any reboot after that (i.e., without
a change) would reset the CPU voltage (and temps) back to their nominal
values (but as always, the CPU voltage setting is still reporting 105%,
110%, or 115%, as appropriate).

My first inclination was to blame a possibly failing PSU, but given this
testing, I doubt it's the culprit. It's a high-end Enermax unit, only 2 ½
years old, and the system is hardly overloaded. I even tried returning the
system to a non OC state (i.e., 133MHz FSB, but it didn't help, same
problems). I'm not even pushing the system or entering Windows, I'm just
rebooting into the BIOS. Prime95 seems to be a victim of this problem, not
the perpetrator.

It appears the BIOS is only recognizing a change in the CPU voltage setting
the FIRST TIME AFTER REBOOT, then seems to "forget" on subsequent reboots,
and applies the default voltage "By CPU" (despite the setting of 105%, 110%,
etc., still being present). I've tried disabling the shutdown temp,
shutdown on CPU fan fail, etc., any number of other settings looking for a
correlation, but can't find one. And I don't feel like running out to buy
another $80 PSU only to find this is a BIOS error.

As one final test, I decided to set the CPU voltage to 110%, but this time
NOT check the BIOS, but simply run Windows and Prime95 immediately after the
change. Now Prime95 runs PERFECTLY! I stop Prime95 after several hours,
reboot (no further BIOS changes), run Windows and Prime95 again, and of
course, it now fails in 10 minutes. Clearly, Prime95 is failing due to the
lack of CPU voltage, but I can't hold any CPU voltage increase past one
reboot (not without updating the BIOS before booting Windows).

Seems obvious to me there's something wrong in the BIOS (BIOS 16 anyway). I
can't understand how a PSU problem could manifest itself ONLY after a second
reboot of a change in some arbitrary BIOS setting. If the PSU was bad, it w
ould ALWAYS be bad, and certainly not in conjunction with an otherwise
completely unrelated BIOS change.

I'd be interested if anyone else has observed similar behavior.

TIA

Jim