CMOS problems on ASUS P5B VM DO

It's been an interesting week. First my cache drive started to fail, so I replaced it. Then, as a result of what I was doing there to troubleshoot and replace it, and some inexplicable lock ups, I noted some odd errors in the syslog. The root cause of that issue seems to have been that when I added RAM several months ago, I failed to notice that it was a different clock / timing than the RAM already present. Memtest gave errors with all the RAM present, but isolating each stick yielded no errors. The end results was that I removed the 2x512MB super talent 533MHz DDR2 sticks and installed only the 2x1GB Patriot 800MHz DDR2 sticks instead of all of them together. Memtest then yield no errors.

Since I was doing all that work, I went ahead and installed the less power hungry Celeron 440 processor that's been sitting around in its box for a few years because I was too timid to mess with any of the hardware Limetech had shipped to me. It's FSB speed matched the upgrade RAM, so it made sense.

Anyway, so now everything seems to run fine. I can set up everything properly in the BIOS, boot into unraid, and have none of the mysterious errors in the syslog. *But* if the system loses power (say, when I try to move it back into the server closet) I get bad CMOS checksum and "overclocking failed" errors. I'm not doing any overclocking. The obvious answer is that the battery is dead, but I've tried 4 different replacements, old and new, all with the same results. So long as power is maintained and the green LED on the motherboard remains lit, I can shutdown or reboot and all is well - but lose power and the CMOS loses it's settings.

My only guess is that I stirred up some dust during all that work and it's lodged somewhere shorting the CMOS, but I've used a lot of canned air to no avail. Updated BIOS to latest version as well. I found this hit of someone having the same problem 3 years ago, but no solution.

http://www.tomshardware.com/forum/216195-30-cmos-checksum-help

Any suggestions?