Re: "[mi] EQ overflowing. The server is probably stuck in an infinite loop."
I think that I overstated the issue that uncachable memory causes, as there are certain conditions when uncachable memory is desireable. In your case, some component on your motherboard likely earmarks 8MB of your system's RAM, so that region is set as being uncachable to keep your cache from corrupting data in that region of memory.
Anyway, having uncachable entries in the MTRR does not cause crashes per se, but when there is buggy code on your system, it is possible for it to trigger race conditions when regions of memory used for programs are marked as being uncachable. Try upgrading your kernel to 188.8.131.52, your nvidia driver to 256.19.29 and your xorg server to 1.9.3. Those are the versions I am using on my system and things have been better since I fixed the MTRR issue.
An alternative that has worked for other people is downgrading xorg server to 1.8.x.