Check the broken USB port and make sure none of the pins are shorting. An intermittent short could cause power-on issues like you describe.

Check your CMOS battery's voltage with a multimeter to ensure it has around 3 volts, it could be a defective new one.

Unplug the case's power button and see if the motherboard still powers on when plugged in; the power button could be faulty.

The chipset is 440BX which does not support high density RAM. Read more here: https://web.archive.org/web/20180103032407/ht … ram_bx_faq.html

Replace the power supply with something more reputable. Something like a Corsair CX430 should work fine.

Use an 80-conductor cable for the IDE to SATA adapter to allow higher UDMA modes.

Thanks for the tips!
Will check them out and report back w/ results.

Towncivilian wrote on 2023-05-27, 19:00:

Unplug the case's power button and see if the motherboard still powers on when plugged in; the power button could be faulty.

Yeah ! also check the bios and make sure "power on after power failure"/"Restore on AC/Power Loss" or similar is not enabled in BIOS

Hi Guys!
I managed to get ahold of some stuff, so I’ve begun testing.
- CMOS: After replacing the CMOS battery with a known good one (3.3 volts), it seams, that things are fine… It still turns on, if BIOS settings are cleared, maybe a feature?
- MEMORY: I purchased some memory but no luck. A single side 64mb stick was no good, eve though it was 4M*64, and I also got 256mb ECC stick, but that was buffered, so it wouldn’t even give beep codes.
- BIOS INFO: It seems that the BIOS veriosn is R1.11.1107… I don’t know if this is the latest tested one, but from the Fujitsu website it seems, that this version might be untested.
- GPU: Cracks on the top of the memory chips top seem to be broken? Could this be an issue? It may only be some top layer for marking and stuff… POST is fine…
- USB: I tried to measure the resistances. No shorts to each other or the ground. I also measured the resistances. I’ve looked around and i think USB should have 15kohm resistors pulling down to ground? Got results consistent with that for the working port. As for the other one I measured… Nothing? What could this mean?
I still need to get some floppies to get going (They are on the attic, hopefully the magnetic stuff is still good). I also have some CDs. What software do you recommend for further testing?

Got myself some more memory, and win 98 is up and running!
Windows seems to run fine, but every now and then it likes to hang. Memtest says that the memory is OK (640 megs).
IDE controllers seem to behave wierdly, and since the USB ports and IDE controllers are on the south bridge... who knows.
Also, something about not being able to control PCI configurations, or an A20 line when shutting down?
Anyways. I guess I have to live with the quirks.

There's been a lot of talk about Win98 bugging out with too much RAM, but I don't have experience with the subject myself. I think 512MB is supposed to be a limit and people generally advise 256MB.

IRQ and hardware addressing resource conflicts can cause those hangs in Win9x. See if you find anything that appears to conflict in device manager. Some devices can share IRQs though, so it's not always a problem. You might have to experiment with disabling/removing unneeded cards to see if that narrows down the problem.

I also got 256mb ECC stick, but that was buffered, so it wouldn’t even give beep codes.

440BX supports ECC and also supports registered/buffered memory. However, registered and unbuffered can't be mixed. If you use this module by itself, it should work, presuming there isn't some other compatibility issue.

Prime95 "torture test" is a good CPU/FSB/RAM stress test. Let it run a few hours, there should be no errors.
If your board allows it, try running that while the FSB is overclocked by 3% above your intended setting. If the setup is stable then it should pass any test with a 3% bus overclock. Then you know it's within safe margins at the setting you're really running it with.

===
You can use your multimeter to check voltages while the system is running. A multimeter can't show high frequency fluctuations (that requires an oscilloscope), but it's still a very useful thing to check on any system IMO.
BIOS health monitor / onboard sensors are not reliable - whether they're accurate is pretty much a coin toss from one board to the next.

Checking 5V and 12V is easy, do this at minimum. You can check 5V (red) and 12V (yellow) on one of the 4-pin drive power connectors. Put the black/negative probe into one of the black terminals or against the PSU case.
Voltage should be steady and preferably at or slightly above 5.0V/12.0V. It shouldn't sag much with load fluctuations.
Officially the spec allows +/-5%, but if you get near the lower end of that you'll probably have issues.

You can also check the battery voltage while system is running - many boards do use it while running. The top side of the battery is positive, just keep the negative probe on the PSU case or a black wire as above.

Checking 3.3V power is a little more tricky, if you want to do that one way is by backprobing the orange wire on the ATX connector. If you push it into the back of the connector you can get it into contact. If you prefer you can do that while system is off, then once the probe is in place turn it on to watch the 3.3V rail as it boots.

CPU Vcore can usually be tested on the large metal tab of a MOSFET near the CPU, but that might be tricky to reach on slot-1 boards. Typically half the MOSFETs in that circuit will have Vcore on the tab, the other half will have 5V there (or 12V on later boards, whatever rail the Vcore power is being drawn from).
If you want to probe voltage somewhere on the board but you're worried about arcing anything - put some tape on the probe leaving just the tip exposed.