suppose you have tried another PSU to rule out faulty PSU?

You missed +5v.
3.3v and -12v are out of spec high.
+12v is barely in spec on the high side.
The PSU is your prime suspect but that may not be the only (or the real) problem.

Bad caps can fail towards shorting internally and one on +5v causing excess +5v current would cause the other rails to go high as the PSU tries to compensate.
If that's the case then +5v should be on the low side if in spec at all.
That makes the PSU try to raise the +5v voltage and the other rails are affected too.
I've seen that exact scenario a number of times.

Look for bloated caps (on the mobo and in the PSU) - but not all bad caps bloat.
If none then try a different PSU.
If the issue doesn't go away look for other problems.

PCBONEZ wrote:

3.3v and -12v are out of spec high.
+12v is barely in spec on the high side.

-12V:
is: -12.3V, allowed range: -10.80 to -13.20V

+12V:
is: 12.5V, allowed range: +11.40 to +12.60V

Both voltages are fine and fully within specs...

+3.3V may not, but most cheap voltmeters are off by 10% or more anyway, especially for very low voltages. So unless he uses a good, calibrated voltmeter, I'd guess the PSU is fine.

derSammler wrote:

-12V: is: -12.3V, allowed range: -10.80 to -13.20V […]
Show full quote

PCBONEZ wrote:

3.3v and -12v are out of spec high.
+12v is barely in spec on the high side.

-12V:
is: -12.3V, allowed range: -10.80 to -13.20V

+12V:
is: 12.5V, allowed range: +11.40 to +12.60V

Both voltages are fine and fully within specs...

+3.3V may not, but most cheap voltmeters are off by 10% or more anyway, especially for very low voltages. So unless he uses a good, calibrated voltmeter, I'd guess the PSU is fine.

Yeah I read -12v wrong but that's for biasing not power so it doesn't matter very much.
A properly working PSU at idle should be in the middle of the range not right at the edge so whether you like it or not those voltages raise red flags.
.

If you're using the BIOS health monitor to read the voltages, then the values aren't reliable. Many boards will give some wild readings that don't match what you'd find with a multimeter.

shamino wrote:

If you're using the BIOS health monitor to read the voltages, then the values aren't reliable. Many boards will give some wild readings that don't match what you'd find with a multimeter.

Yes - This.

Wow, sorry for being late with my reply. I didn't get any answers in the first couple days, so I gave up on it.

I have, in fact, checked it with a different PSU, but that was also on a backup motherboard. Didn't get any 3DMark crashes there, but the horizontal flicker was still kind of there. I now have a new WinFast GeForce2 Pro which doesn't flicker, but 3DMark still crashes. I suppose there are some bad caps hiding on the motherboard, as it behaves somewhat oddly in certain aspects (sometimes the boot memory test lags and the optical drive isn't detected by the BIOS after soft resets).

Also, the hardware monitor in the P2B's BIOS is pretty accurate, except for the CPU VCore which displays 2.1V. Using CPUID HWMonitor, I was able to determine that the 3.3V rail does indeed operate at 3.63-3.65 volts and the rest of the values are also as shown by the BIOS, unless the CPU running at 2.06V most of the time rather than 2.05 counts as unusual.

PKFreeZZy wrote:

Also, the hardware monitor in the P2B's BIOS is pretty accurate, except for the CPU VCore which displays 2.1V. Using CPUID HWMonitor, I was able to determine that the 3.3V rail does indeed operate at 3.63-3.65 volts and the rest of the values are also as shown by the BIOS, unless the CPU running at 2.06V most of the time rather than 2.05 counts as unusual.

Software utilities still read from the same sensors that the BIOS does, so they will say the same thing (barring a shift due to system load, or because the software isn't interpreting the data correctly). But the accuracy of motherboard sensors is generally poor, sometimes they're close but sometimes they're way off. There's no way to know without an independent measurement.
The 3.6V reading in particular is concerning so I'd want to verify it with a multimeter. If it really is 3.6V then I'd keep any valuable hardware away from that setup.
+5V might be low to the same degree that 3.3V is high, as PCBonez alluded to. Some PSUs are group regulated in a way that causes this to happen.

5V(red) and 12V(yellow) can be measured on an unused drive power connector. 3.3V can be backprobed on the ATX motherboard power connector (orange wire).

Well then, I guess I don't need to be concerned about the PSU; I know for a fact it operates properly because I checked it on a modern LGA1156 motherboard and the 3.3V line delivered 3.392 on it; didn't know the 5V rail is also almost out of spec, but my the other systems I've used with this PSU all worked flawlessly.

Could this really be caused by the capacitors? They seem fine, there's no leaked electrolyte nor any bulging ones, though I've learned my lesson with the "backup motherboard" I mentioned as that one sometimes behaves strangely as well and its caps also look fine.

I've considered replacing the ones on the P2B with polymer capacitors since I think there's no other source of trouble than ancient Cylinder Town.