It is certain that the CMOS clear jumper is in the right position ?

I would say so, yes. If it wasn’t, then it would reset at a normal shutdown or reboot, no?

Actually yeah, it should stay in cleared state then and not store settings at all.

Your description strongly suggests that there is only +5VSB connection going but nothing from battery. I have actually seen a fault like that on one socket7 board, where if CMOS clear jumper is set during operation it'll burn a trace, as that jumper directly shorts the CMOS power rail to ground and during operation it is given full 5V rather than weak current from battery. Only reason the board held any settings (as long as there was power) was through ESD diodes or something such inside the chipset. On that particular board all I needed to do was patch a trace. Do you have a multimeter ? If you can follow the positive trace from battery around the board a little you may be able to find a bad passive etc. if this board implements things the same way...

Ah. That’s way above what I can do, unfortunately. I think I’ll live with it and hope I don’t need to cut the power unnecessarily. 😀

I have a super micro board that where the pins from the battery had bad solder joint or something and it wouldn't hold settings. I just had to reflow the 2 joints on the backside of the board and it has held settings ever since.

I had an issue like this once - the solution was that there was some corrosion on the coin cell holder's contacts. It wasn't easy to see, I only figured it out by checking the voltage on the back of the motherboard where the battery holder's contacts are.

The contacts on the holder looks brand new and there is no corrosion. What bothers me is that it only resets when the power is cut of unexpectedly. A normal shutdown or reboot does not reset the cmos. So I really can’t see it’s battery related. But it’s an old setup so I guess there is a fault somewhere on the board itself. 🤷‍♂️

A normal shutdown won't cut the power, the PSU continues provides 5V StandBy voltage as long as it is physically plugged to the wall and is turned on from any switch on it, which the board uses so that the battery wouldn't be drained unecessarily. Reboots don't cause anything to the power bits either.

Arnull wrote on 2024-01-31, 20:13:

The contacts on the holder looks brand new and there is no corrosion. What bothers me is that it only resets when the power is cut of unexpectedly. A normal shutdown or reboot does not reset the cmos. So I really can’t see it’s battery related. But it’s an old setup so I guess there is a fault somewhere on the board itself. 🤷‍♂️

If it were me, I would make sure that you see 3 volts when measured across the contacts on the back of the board. If yes, I'd move on to examining the solder joints for the battery holder, then following the trace from there.

I also once had a Chaintech Super Seven AT board that would lose the CMOS settings whenever I moved it. I finally realized that there was a motherboard standoff with no corresponding hole - moving the computer would ground the motherboard and clear the CMOS memory. It was my own fault for not being more careful when I assembled it.

Well, new IDE cables seem to make the prolem go away. Go figure. 😀