I'd suggest you get an 'old school' 3xAA battery holder and use that instead. I've never had luck with either a single CR2032 or even a dual CR2032 holder working correctly on any of my 486 motherboards. Latest failure there was a Micronics JX30G, tried a dual CR2032 holder and the BIOS time would lose an hour a day and within a month it quit retaining settings. Switched to a 3xAA holder and everything has been perfect since. Just my two cents.......

386DX40 wrote on 2026-02-14, 15:08:

I'd suggest you get an 'old school' 3xAA battery holder and use that instead. I've never had luck with either a single CR2032 or even a dual CR2032 holder working correctly on any of my 486 motherboards. Latest failure there was a Micronics JX30G, tried a dual CR2032 holder and the BIOS time would lose an hour a day and within a month it quit retaining settings. Switched to a 3xAA holder and everything has been perfect since. Just my two cents.......

I'm so glad you said that. One of my thoughts was that the voltage diff between the barrel (3.6) and the button cell (3.0) was too great, especially since the readings probably change when under load, even a load as small as what is needed to keep those settings. Thanks so much, I think I'll try a set of NiMh AAs fully charged to see if the behavior changes. If it works I can solder it onto the headers where the barrel sat and they should get a small charge as well.

Funny thing, AI suggested a cordless phone battery pack, as the voltage is closer and the design of the pack allegedly helped fight the off power shelf drain and long discharge times (like if someone didn't return it to its charging base when they finished using it).

I'll try it out and let you know. Thanks again!

I have seen boards that come with CR2032 do that when the battery starts going flat. It seems as if the RTC part needs more volts to keep ticking than what the memory part needs to retain settings. Perhaps the 3.0V is just a tad too low for those RTCs to tick.

Just to confirm: if you go into the BIOS when the systems are on, the clock is seen running/ticking?

st31276a wrote on 2026-02-14, 15:52:

I have seen boards that come with CR2032 do that when the battery starts going flat. It seems as if the RTC part needs more volts to keep ticking than what the memory part needs to retain settings. Perhaps the 3.0V is just a tad too low for those RTCs to tick.

That's really interesting, thanks for sharing. Over half a volt difference at nominal seems excessive to be deemed 'okay'... I'll see if I can pick up some cells locally today and run my test.

jmarsh wrote on 2026-02-14, 16:23:

Just to confirm: if you go into the BIOS when the systems are on, the clock is seen running/ticking?

Yes, the clock is ticking normally. The time and date are about where they were when the power was last shut off. (i.e. I turned the second board on last night after setting the clock about 24 hours beforehand, and it still showed a few minutes past 9pm on the 13th, but was ticking normally.

3V is not enough for this board when using an external battery. I've had the same issue when using a CR2032 or 2xAA. Go with 3xAA or 3xAAA instead.

The onboard 3.6V barrel works fine because usually the internal battery goes through a path with a smaller voltage drop than the external header.

Thanks for your input everyone! I got a set of new Energizer NiMh AA batteries and a new battery holder which read just over 4 volts after a full charge. I plugged the holder into the external battery pins and set the time. It held time overnight and is in line with my watch. That was it, the CR2032 wasn't enough.

Big relief as I thought there were bigger issues initially. You all saved the day!

I have a question for the mavens out there... I have also experienced the same issue as the OP and performed the same fix (using a battery holder with lithium AAs instead).

But I also swapped out the UV EPROM with an EEPROM. Do ceramic UV EPROMs consume more battery?

The BIOS ROM should be off when the system is powered down, only the RTC chip should be powered.