Pretty much since the 915P chipset, AUTO on the PCI-e freq seems to have worked best rather than trying to lock it at 100 (at least on ASUS boards).
What are you checking actual PCI-e MHz with? I am curious because there is/was PCI bus "float" problem on the 915 boards. The "lock" didn't really lock it. It held it lower than it would unlocked, but it would still drift up with increased FSB and if you're seeing 108MHz on it at 240FSB, I wonder if that float issue has somehow been carried forward to the 945? If you lock it at 100MHz in BIOS and boot up @ 240FSB, what is the PCI-e speed? If the problem is the same as the 915, it will be the same 108MHz or faster and I'll bet $$ that is where ALL your problems are coming from.
915P had the problem, 925XE didn't. Likewise, maybe the 945 has it and the 955 doesn't. My 955 on AUTO PCI-e, it doesn't budge off of 100MHz all the way up to a 400FSB (as high as I can possibly go right now
).
If you have that float problem, your OC is going to be limited and there's nothing you can do about it because the higher you up the FSB, the more out of whack the bus gets and the more out of whack the bus gets, the more unstable the system becomes. With a float, you are basically OCing EVERYTHING on the bus from drives to the vid card. Serial drives in particular are very susceptible to bus changes and will probably be the first things to cause a problem if the bus gets too high. That's why the bus locks are needed in the first place.
So, if the PCI-e is floating, you were correct originally about their being a chipset problem, but the problem is not in relation to the memory control. If it floats and you can verify that it won't stay at 100MHz either on AUTO or set to 100 in BIOS with increased FSB, get rid of that board ASAP and replace it with a 955 (or wait for the 975) or you can forget about OCing much or any higher than you are right now