Re: "[mi] EQ overflowing. The server is probably stuck in an infinite loop."
Posting a followup: Unfortunately, the intermittent errors returned. Neither the kernel MTRR option nor the nvidia-settings application suggestions seemed to have done any measurable good.
I'm in the process of doing an emerge world with the -g flag enabled to see if we can narrow down where it might be triggered from.
I'm also purchasing a new motherboard/cpu/memory combo to see if maybe this affects a certain hardware combination.
Currently: 2 Quad-core Opteron 2378s, 16G of DDR2-667 ECC Buffered RAM, 2 Nvidia GTX 295s (in SLI configuration), 4 1TB drives in mixed raid configurations, and a Supermicro H8DA8-2 motherboard.
Soon to be: 1 Athlon Phenom II X6 (3.2G), 16G of DDR3-1066 Unbuffered non-ECC RAM, 2 Nvidia GTX 295s (in SLI configuration), 4 1TB drives in mixed raid configurations, and an ASUS M4N98TD EVO.
I can at least, at that point, determine if it is hardware (with the drives/video cards being the only constant elements) or a software one (nvidia drivers).
I should have a definitive answer as to whether this corrects the issues mid-next week when the components arrive and I can assemble them. Though, based on what I've read in this thread, it seems that folks in the Intel camp (i7s in that case) seem to have a similar problem. So, it makes me less convinced this hardware combination will necessarily solve the problem so much as bring me more up to date.
Kernel recompilation, especially with unnecessary drivers and features removed, doesn't seem to have addressed the problem. Nor has incrementing to a more recent kernel. The only thing that seems to make my system not spinlock to the point where I can get any data is using a ck kernel using the BFS cpu scheduler. It seems to at least pass the issue around the cpus more evenly so I can gather data and reboot the system more cleanly.