I have the same problem with my P55A UD3 board, latest F8 BIOS, with two 2GB sticks of 1333MHz Corsair XMS RAM, and with only 2GB usable, except rather weirdly, this only happens on first boot up. After that, it shows and uses the full 4GB absolutely fine. If I turn off, and cycle the power switch, it'll still show the full 4GB. If I leave it about 7 hours though, the first boot will again show 2GB usable.
I swapped the sticks, and in the primary dual channel slots, it was still showing 2GB usuable on first boot, except it reported the enabled 2GB for the wrong slot (an empty one??). If I put the sticks in the channel A slots only (the two closest to the CPU), with the one I suspect bad in the closest slot, it has an error message after post about the system experiencing boot failures due to overclocking etc, which I haven't done.
Last night, I tried with the suspected 'good' stick in white slot closest to the CPU by itself, to let everything warm up - then put the other stick (which would have still been 'cold') in the other slot after a few hours use - that still resulted in only 2GB detected on the first boot at POST, so I'm actually thinking could I rule out the CPU and bent pins? Surely in this literal 'cold boot' only situation, if it was a bad cpu contact resolved by heat, then the other stick of RAM should have been detected first time? Again, coming out of BIOS without saving any settings, it shows the full 4GB.
I've got further testing to do, but do suspect the same stick for not showing up as being 'bad'. However, because it has never caused any BSOD's (yet) and I've played Crysis maxed out with custom mods, that resulted in about 3700Mb usage at one point (lol) then I can hardly call it completely bad. I've ran the windows memory diagnostic thing, just on standard runs though, and that passed fine too. Tried each stick individually and together.
The RAM had initially been left at 'auto' voltages, which resulted in 1.584, so it' has been getting plenty of volts too, and I've since lowered that by manually setting 1.5v, which results in 1.536v actual, and that has been 'ok' apart from the same issue too.
Because of this being an issue only when the machine has been turned off for at least several hours, it makes it hard to test things, as I have to wait so long before I try any new settings or slots.
Any ideas much appreciated, and I can provide more info if needed, as I have carried out a few other tests that I haven't mentioned here. I really don't know whether the problem is the ram, motherboard or cpu - I'm hoping it's just RAM, but then i'd be worried to RMA it just in case they can't replicate the problem. I'm reluctant to undo the cooler, and will only do that as a last resort, as I'm getting great temps, and I'm worried about upsetting things when it might not be the cause.