What an odd problem.

Iirc normal mode completes the loop to the battery.

I don't think so, as removing the battery would require you to wait a while for CMOS to completely clear. Setting the CMOS mode jumper to clear mode should work instantly. And I have tested this on several different systems; setting the jumper to clear mode makes the system unable to turn on at all, and the PSU fan won't even spin. This doesn't happen if you just remove the battery. It's of course possible this behavior differs from motherboard to motherboard, and it probably does, but it tells me that it is somewhat normal.

Maybe it shorts a power rail to ground (or connects a low-resistance resistor) to discharge capacitors.

Clear CMOS jumper definitely does NOT short any power rail to ground. If it did, the thin traces that (usually) lead to it would melt in a second before the PSU has any chance to trigger its short-circuit protection.

What you experienced is probably just a motherboard that refuses to start without a CMOS battery. I've ran into a fair share of these last week when testing various retro boards (as that's what a Clear CMOS jumper usually does - it removes power from the CMOS Vcc pin.) A few did turn On and worked fine, but a good few also acted completely dead and/or just "bliped" with power for a short second and then turned Off. I even tried forcing one of these mobos to turn On (thinking it was defective) by jumpering the PS_ON wire on the PSU to ground. All that did is extend the "blip" of power from the motherboard for a little longer, while the PSU did remain running (as instructed by the grounded PS_ON signal line.) In the end, that one turned out to be a bad CMOS battery connector. I had a fresh battery in there, thinking all was well. Not sure how I got to it in the end, but I found the CMOS battery wasn't making contact well. Once I corrected that, motherboard "magically" turned on and worked fine afterwards.

momaka wrote on 2024-11-26, 21:59:

as that's what a Clear CMOS jumper usually does - it removes power from the CMOS Vcc pin

That depends on what sort of RTC the board has.
For the Dallas 12885 style (which includes the DS12887 because it's the same thing with a battery and crystal glued to the top) the clear CMOS jumper shorts the RCLR# pin to GND. The pin is meant to be pulled up internally but some board designers also made the mistake of connecting it to VCC...

momaka wrote on 2024-11-26, 21:59:

Clear CMOS jumper definitely does NOT short any power rail to ground.

yes, more precisely depending on design it can short Vbat to ground _after_ the current limiting resistor. Without explicit reset or shorting vBat RTC could keep its contents for quit a while.

I just realized Iv seen advice to pop the battery and short its terminals a few times - that does nothing because there is a diode after battery holder.

QDI Titanium IB will smoke a trace off the board if it is powered while CMOS clear jumper is set to the clear position, I have had to repair two such instances. Not a direct short, since there's some circuitry between 5V and the grounded battery line, but enough to burn out a trace... I have never seen such a thing happen otherwise though.

momaka wrote on 2024-11-26, 21:59:

Clear CMOS jumper definitely does NOT short any power rail to ground. If it did, the thin traces that (usually) lead to it would melt in a second before the PSU has any chance to trigger its short-circuit protection.

What you experienced is probably just a motherboard that refuses to start without a CMOS battery. I've ran into a fair share of these last week when testing various retro boards (as that's what a Clear CMOS jumper usually does - it removes power from the CMOS Vcc pin.) A few did turn On and worked fine, but a good few also acted completely dead and/or just "bliped" with power for a short second and then turned Off. I even tried forcing one of these mobos to turn On (thinking it was defective) by jumpering the PS_ON wire on the PSU to ground. All that did is extend the "blip" of power from the motherboard for a little longer, while the PSU did remain running (as instructed by the grounded PS_ON signal line.) In the end, that one turned out to be a bad CMOS battery connector. I had a fresh battery in there, thinking all was well. Not sure how I got to it in the end, but I found the CMOS battery wasn't making contact well. Once I corrected that, motherboard "magically" turned on and worked fine afterwards.

Nope, it starts up just fine without a battery. It definitely trips the short protection in the PSU when in clear mode. I can hear it. The PSU fan doesn't start, and when I remove the power I hear a coil whine sound after a while. This is a classic behavior for a PSU in short-circuit protection mode. And I've seen this behavior on at least 3 other motherboards, with different PSUs as well.

Anyway, I'm not here to fight, but I wanted to share my story with other people in case it could help them out as well! In my case, removing the jumper is the same as putting it in position 2-3 (clear mode), so a bad connection between the pins in normal mode was enough to cause this issue.