This problem is very strange. The actual cause is unknown, but when it happens, it seems to be impacted by the voltage supplied to the 735 chip and/or the temperature of the 735 chip.
Click here for the full FAQ on this problem.
The common symptoms:
1) BIOS 2nd screen: Lists "DDR DIMMs: x y z" where there are more numbers (x y or z) listed than DIMM's installed
2) Windows 98/Me: CONFIGMG errors when starting Windows
3) Windows 2000/XP: Error messages about missing or corrupt files when starting Windows
4) CMOS settings suddenly lost for no reason
Current work-around and "fix" options:
1) Add a 470 Ohm resistor in parallel to R182. This will lower the chipset voltage slightly (about 4-5%) and has fixed the problem on many boards. (Because each 470 only lowers the voltage by 4-5 percent, you might need to add a second resistor if problems continue.)
Method 1: Use a 470 Ohm (1/4W 5% recommended) resistor (Radio Shack) and solder it between the left pin on U8 and the 3nd pin of the LAN LED connector.
This photo shows a 1/4 watt leaded resistor installed. Click here to see a photo of a resistor attached to an alternate grounding point (for non-LAN boards). (Photo courtesy of Oken.)
Method 2: If you want a neat looking fix, use a 470 Ohm surface mount resistor instead and apply it right on top of R182.
See here for a photo of the location of R182 where you can solder a surface mount resistor on top.
Note: I have no solid idea why work-around #1 helps, and it may turn out that this is a very strange way to fix the problem, if the real cause of the problem is ever discovered.
If you can't do 1), you can consider the following:
2) Overclock the FSB. (Use the OC BIOS or the chfsb utility and run 138/138 or higher.) This has helped reduce the problem on a number of boards. It is thought that this helps sometimes because it makes the 735 chip run hotter.
3) Turn the option for "Quick Boot" off in the BIOS. This has helped a number of boards. It this thought that this helps sometimes because it allows the 735 to get a little hotter before booting.
4) Remove the 735 heatsink. This lets the chipset run hotter, which can reduce the problem in some cases.
5) Cover the 735 heatsink with some thin antistatic foam (cut a chunk from the sheet that the motherboard came in) to let it get a little warmer.
6) If options 1-5 don't work, keep the 735 extra cool by putting a good heatsink on it with a fan.
More about the problem:
If the voltage is around 1.7V to 1.75V for the 735: There never seems to be a problem.
If the voltage is 1.8V to 1.85V, some (few) boards can have the problem.
When the problem happens, it is often only present at specific 735 chip temperatures. It is not just when the 735 is too hot, or just when too cold. It seems to happen when the 735 is in a range of temperatures.... If it's cold, it's OK, if it's hot, it's OK. In the middle, it's no good!
With the voltage too high, different boards have different temperatures that have the problem. The temperatures might be from 16C to 20C, or from 27C to 40C, or from 30C to 40C, etc. Unfortunately, it's almost impossible to make sure your chip is never in the temperature range because the temperature of the room can change, as well as the amount of time the computer is on.
By only controlling the 735 chipset temperature, it's usually possible to get the board to work well enough to get by, but until the voltage is lowered, the "Lost CMOS" problem seems to continue.
Therefore, the recommended work-around is the 470 resistor added to R182. It is not known if this is the best solution, but it has worked on 100% of the boards tested, and, so far, the boards remain 100% solid in testing.
Because it takes a long time to know if your "lost CMOS" problem is really gone, it is still hard to say if work-around #1 is a complete fix. So far, fixed boards that suffered from "lost CMOS" almost never since shown the problem. (In one case, one board out of the many I've fixed said the CMOS was lost after about 3 days of total power-off.)