Sounds like something is overheating or there is a bad solder joint somewhere.

I aggree that the symptoms suggest overheating. I did some microscopic observations regarding solder connections around the major chips but they all look pristine. How would I do a more systematic troubleshooting (I dont have an IR camera). Which parts would be most prone to cause the observed behaviour (e.g. could it be the CPU ?) - thanks

Aui wrote on 2025-02-27, 01:55:

I aggree that the symptoms suggest overheating. I did some microscopic observations regarding solder connections around the major chips but they all look pristine. How would I do a more systematic troubleshooting (I dont have an IR camera). Which parts would be most prone to cause the observed behaviour (e.g. could it be the CPU ?) - thanks

if you had a heat gun you could start heating isolated areas of the board and see if that can speed up the crash

start with the usual suspects ... the chipset, cpu, and then 47 glue logic

And is that big round black thing you CPU cooling solution ?
DX4 100 does get quite hot quite quickly ... maybe a setup with fan could exclude the CPU from this issue ?

Good luck

Nunoalex wrote on 2025-02-27, 13:38:

And is that big round black thing you CPU cooling solution ?
DX4 100 does get quite hot quite quickly ... maybe a setup with fan could exclude the CPU from this issue ?

Good luck

Agree ! re-do the HS and use some thermal compound grease.

Thanks for the suggesions - I aggree the heat sink solution is odd (very strange design) and maybe it damaged the CPU over time ?. However, I tried with a much larger (better) heatsink and the system crashes before the cpu gets even warm. I have a spare SX25 which could be used for testing but I am uncertain how to configure the board / FSB. I only see a 3 - 5 V jumper but all other jumpers are just numbers and I dont have a documentation (I attach a few more images with the jumer areas).

There is also some evidence that the problem may be related to the graphic (chip ?). Sometimes the system sems to be allive without VGA output. E.g. I can get systematic HDD activity when starting a program while the screes is already black. Sometimes starting the system wakes up the monitor without showing anything? The S3 chip also only gets mildy warm - nothing serious

Well, reseat every socketed chip for starters.

Feel every chip on the board when it crashes and see if any of them are oddly hot.

It also could be a capacitor somewhere.

Ok - I reseated socketed chips - there is only the Bios and the two long slim bar shaped chips (dont know what they are) left of the coincell that can be reseated - done but no effect. All other chips are soldered to the board. None of the chips gets very hot - rather all chips sem to get to about 20 degree.

About thoses capacitors - there are only 2 classic electrolytic ones. I probably could change those - is this worth a try given the current failure behaviour ?
There are also a number of smaller barel shaped (polymer ?) capacitors for example close to the power connector which are all surface mounted (daunting - have never recapped those). Would they fail visibly ?
Plus there is a huge number of tiny caps (small 1 -2 mm rectangular blocks). Also the backside of the board houses a large number of those. They all look good. Would those visibly fail ?

Most caps don't visibly fail unless the electrolyte physically leaks out the bottom or bulges out the top - but they often fail regardless of their appearance.

The surface mounted ones can sometimes leak underneath until they corrode their own terminals and are easily snapped off with almost no force.

Don't worry about the tiny ones. Those are not electrolytics and probably won't fail.

Your issue is most likely either a bad solder joint or a flaky silicon junction in a chip somewhere, unfortunately, but it is worth it to try to grab low hanging fruit like bad electrolytics if you can't do some more in-depth diagnostics.

I will say this though, in the third pic in your OP, the large electrolytic by the CPU socket is bulging a little on the top.

I would check for loose legs on the chipset too. One of the chips is very close to the CPU and the temperature levels were probably higher in that area. This could speed up solder aging and break some connections.

Aui wrote on 2025-02-28, 04:51:

Ok - I reseated socketed chips - there is only the Bios and the two long slim bar shaped chips (dont know what they are) left of […]
Show full quote

Ok - I reseated socketed chips - there is only the Bios and the two long slim bar shaped chips (dont know what they are) left of the coincell that can be reseated - done but no effect. All other chips are soldered to the board. None of the chips gets very hot - rather all chips sem to get to about 20 degree.

About thoses capacitors - there are only 2 classic electrolytic ones. I probably could change those - is this worth a try given the current failure behaviour ?
There are also a number of smaller barel shaped (polymer ?) capacitors for example close to the power connector which are all surface mounted (daunting - have never recapped those). Would they fail visibly ?
Plus there is a huge number of tiny caps (small 1 -2 mm rectangular blocks). Also the backside of the board houses a large number of those. They all look good. Would those visibly fail ?

The Toshiba T4400C laptops I like trying to fix use those same Elna caps, which are the main failure point on that laptop. The capacitors don't fail visibly from the top but they can leave signs of leaked electrolyte when removed and on a couple of boards that electrolyte has damaged soldermask and traces underneath.
The Elna Longlife capacitors generally seem to be holding up, but the regular brown elna capacitors do tend to leak in my experience, or lose capacity or go high ESR. The inductor by the CPU suggests that it's a switching power supply giving 3.4v so the big capacitors near the CPU could actually have a tough job and really need to be in spec for the system to work.

If cooling isn't found to be the issue, that's where I'd look next.

Additionally, the SMD capacitors by the AT power connector, the solder pads look discoloured / matte rather than shiny. I'm not sure if that's just the angle the picture was taken from but if those solder pads are matte / grey instead of shiny silver then it's possible those are leaking and causing issues with traces beneath.

edit: ehh, maybe not all Elna caps are bad and perhaps there's just a certain date range or capacity that's more affected than others. Just pulled some 6.3v 1500uF Elna (RCX) caps with a date code of week 18 1995 and they all test good

Ok - I will start with the two electrolytic ones close to the CPU. Will need to order them first and then report back. Thanks for the sugestions!

Aui wrote on 2025-02-28, 04:51:

There are also a number of smaller barel shaped (polymer ?) capacitors for example close to the power connector which are all surface mounted (daunting - have never recapped those). Would they fail visibly ?
Plus there is a huge number of tiny caps (small 1 -2 mm rectangular blocks). Also the backside of the board houses a large number of those. They all look good. Would those visibly fail ?

Those barrel shaped ones are electrolytic too. I'm not sure how prone they are of failing though. The small rectangular ones are ceramic, they don't fail unless mechanically damaged ..

STABLE !

Changed the two electrolytic ones close to the CPU - and the system is running rock solid!

Thanks for encouraging sugestions and comments !

You are lucky. Glad it worked.

Glad to see you made it work, who doesn't like a good happy ending 😀.

Aui wrote on 2025-03-06, 06:35:

STABLE !

Changed the two electrolytic ones close to the CPU - and the system is running rock solid!

Thanks for encouraging sugestions and comments !

And the two didn't look bad at naked eye..

Aui wrote on 2025-03-06, 06:35:

STABLE !

Changed the two electrolytic ones close to the CPU - and the system is running rock solid!

Thanks for encouraging sugestions and comments !

It is never caps until it is CAAAAAAAAAAAAAAAAAAAAAAAAAPPPPPPPPPPPZZZZZZZZZZZZZZ 😀