The southbridge contains the real time clock, ATX start and shutdown circuits ect. so it´s hooked to the 5V standby rail of the PSU.
There are several PSU models that don´t provide sufficient protection against a too high standby voltage in case of a PSU defect. It´s quite possible the faulty standby voltage damaged the southbridge.
The "ground issue" could be due to some burned ground traces on the mainboard. You should inspect both sides very carefully with a good magnifying glass.

will 100 degrees celsius kill a chip? my router's cpu can run at 90 sustained on a hot day.

do you have a psu tester? it sounds like your replacement psu is defective. it's happened to me before. i had a string of 4 in a row go bad due to 5v rail draws (athlon cpu).

Unfortunately, the southbridge appears to be burned out...
See the image...

You need to open up and inspect all old hardware, including (or especially) PSUs.

Incidentally, the most dust I've ever seen in one location was in an old IBM XT PSU. Literally every empty void in the case was packed solid with dust. It was incredible. I dread to think what would have happened if I had plugged it in without cleaning it. (It would probably have caught fire.)

Out of interest what PSU was the killer in this case?

RIP, never trust old ageing power supplies, their capacitors are generally on their last legs and they will go off and can cause the power supply to die and take other hardware with it.

Though the motherboard caps were also too old to trust whether they looked good or not and may have contributed. A fully good PSU might have detected problems and shut itself off, a fully good motherboard might have dropped power good if the PSU looked flaky. So likely a bit of both IMO.

stanwebber wrote on 2022-12-13, 08:51:

will 100 degrees celsius kill a chip? my router's cpu can run at 90 sustained on a hot day.

Apart from chips that come in ranges like 74 series or similar, the Tmax in operation is specified differently for each part, though some may be the same. You can see as high as 120 celsius and as low as 60. AMD typically have temperature codes in their part codes which can be looked up. There's not that many below 70C though so that's a safe level to attempt to hold to unless you have specific info about super sensitive ones. Max storage and soldering temps are different. Storage assuming they might be at that for a long time, soldering assuming they will be at that just a few minutes at most.

It's not necessarily about the PSU being old, it has more to do with the setup being from the early '00s (meaning capacitor plague) + a poor design of the standby circuit which is known to kill motherboards once certain critical caps fail. This is the real danger of ATX PSUs of that era, usually stuff that is powered by the +5VSB line tends to die (southbridge, memory and so on). A properly designed power supply can fail when its time comes like everything else, but it won't kill other things unless there's a catastrophic failure.

Still, the OP has not provided details about the PSU yet, so it's all speculation at this point.

I wish everyone would forget about the plague like it labels only some caps as bad, ALL CAPS GO BAD eventually, and here we are, 10 years past expected lifetime. A decade back it was minorly mattering whether caps were from plague manufacturers or not as to whether they had a couple of years left in. Now, forget it, everything has equal chance of being bad. Otherwise it's like "Here's some 2 year old milk from a dairy that had old pasteurisation equipment so they only put a week sellby date in it, and heres some from that had up to date at the time equipment and could put a 2 week sellby date on it, also 2 years old, which do you want to drink?" Neither they're both too freaking old to be good.

And that's the reason it's dumb to complain on auction sites about old hardware failing even though it was tested earlier and looked fine.
No wonder people sell these AS-IS on Ebay, imagine getting negs for stuff that was made like 30 years ago but buyers still expect them to work like they are brand new.
When you buy retro you have to accept the fact that many of those won't work at all or just blow up the second you power it up.

No, I don't agree, there's plenty of stuff (even much older) that is working just fine and the old caps are well within spec even today. The milk analogy is not good, electrolytics don't work like that and quality ones last much longer than what people give them credit for, especially if they've not been subjected to lots of heat. The lifetime ratings from the manufacturers are only provided under extreme conditions (max. temp and max. ripple).

Capacitors affected by the plague in hardware from 2001 to 2005 have way more chances of being bad, so much so that some bloat and leak even without having power applied since they left the factory.

Besides, my point was more about the hardware being killed by the bad design of the standby line in the power supply.

Mandrew wrote on 2022-12-13, 16:38:

And that's the reason it's dumb to complain on auction sites about old hardware failing even though it was tested earlier and looked fine.
No wonder people sell these AS-IS on Ebay, imagine getting negs for stuff that was made like 30 years ago but buyers still expect them to work like they are brand new.
When you buy retro you have to accept the fact that many of those won't work at all or just blow up the second you power it up.

I don't necessarily agree with this as the sellers often aren't entirely forthcoming (or just outright lazy). If a seller is selling a vintage PC showing it as working then they really should "do their best" and photograph the motherboard, try to look at the power supply internals and document anything they see questionable. Often they want to "dump" these older units under the guise of as-is no warranty (but it posts!), yet charge what other sold items go for from a seller who has actually done this work.

Anyone getting into retro may not be aware of everything and might assume that a BIOS screen post shows proper functionality. I've seen some very bad/leaky caps on motherboards and the PC still posts just fine. But it'll die pretty quick if pushed.

To add to this...most of the caps in my vintage audio receivers (40+ years old now) are still perfectly fine and still in spec. Those older quality caps didn't leak or explode...they just went out of spec.

Even something as recent as a 486 I have all of the caps both on the mobo and in the power supply are just fine.

_tk wrote on 2022-12-13, 17:00:

Often they want to "dump" these older units under the guise of as-is no warranty (but it posts!), yet charge what other sold items go for from a seller who has actually done this work.

While I agree with this I also think it's ridiculous to expect warranty for something this old/delecate even if it was tested a week before. This post is ALMOST proof: a PSU that was on the verge of failure crapped out unexpectedly even though it was probably fine before. Caps reforming, suddenly shorting, spilling their guts after 30 years of abuse/poor storage, shipping damage, condensation and we could go on.
I've learned that nothing's certain with this hobby and I've never complained about old stuff not working even if it was supposedly tested before. I actually like untested more because it often costs considerably less, they are from scrappers who don't have the time or the will to test them one by one. It's a lottery but whatever, adds to the excitement and an excellent opportunity to get into repairs. Whatever floats your boat.

_tk wrote on 2022-12-13, 17:04:

To add to this...most of the caps in my vintage audio receivers (40+ years old now) are still perfectly fine and still in spec. Those older quality caps didn't leak or explode...they just went out of spec.

Even something as recent as a 486 I have all of the caps both on the mobo and in the power supply are just fine.

I think - besides bad quality and operating out of specs due to sub-optimal design - the main reasons for ageing / failing electrolytics are
- high temperature
- high currents / ripple currents
- high frequencies

All this didn´t matter too much in the age of (now) vintage audio equipment and early PC mainboards and PSUs.

Around the time when Athlons and Pentium 4 CPUs got introduced all these parameters increased and the lifespan of electrolytics in the respective circuits (PSU, VRMs) decreased rapidly.
To make things worse the capacitor plague was also still an issue during those years.

I have an IBM P4 with all the capacitor leakade on the mainboard and work perfect, probably the mainboard design is very good

I guess these systems, at least the quality ones, are built with plenty of redundancy, and will continue to work even if some of the capacitors malfunction or fail.

TheMobRules wrote on 2022-12-13, 16:49:

No, I don't agree, there's plenty of stuff (even much older) that is working just fine and the old caps are well within spec even today. The milk analogy is not good, electrolytics don't work like that and quality ones last much longer than what people give them credit for, especially if they've not been subjected to lots of heat. The lifetime ratings from the manufacturers are only provided under extreme conditions (max. temp and max. ripple).

Unless you've got a crystal ball you've no idea whether all the lifetime got used up in 24/7 on time for 5 years, or whether they had a "perfect for preservation" life of being turned on every 2 months to present day and ran for only a couple of hours. Yes you can lab experiment a "better caps are better" scenario, but picking up stuff that's been out in the real world and had untrackable use cycles, total crap shoot. The odds are on the side of old caps being bad.

This has been mentioned before, but certain ATX PSU designs turn them into literal motherboard killers when caps on the 5V standby rail go bad.

See Re: A tale of two PSUs