So recently, I swapped motherboard/cpu combo's from my primary rig and my GF's rig. I swapped my Asus sabertooth P67 & 3570K with a EVGA P67 FTW & 2500K. Mainly for 3-way SLI possibilities & the overclocking ability of the 2500K(my 3570K hits A NASTY wall 4.4Ghz & recently lost stability@4.4Ghz and is starting to require more voltage). Anyway I had run some benchmarks before I had done the swap with my 3570K@4.2Ghz with the Sabertooth board & then run with the new board with the 2500K@4.6Ghz. Comparing my results before & after is interesting given the faqs. My physics score has gone up nearly 400 points, however in all the graphics tests, I have lost 2-3FPS. What i find interesting is with the Sabertooth board in SLI only runs PCI-Eslots at 8x8, where as the EVGA board is running 16x8, so it should have a decent amount more bandwidth if needed. With the bandwidth increase & a faster CPU to further eliminate CPU bottleneck, I was for sure expecting a slight increase. To add some complication to this story & against my better judgement, I did NOT Format! Being as it was the same chipset, I decided I'd see if I could get things running smooth, as I have in the past many times wtih no fuss. And so far I've had zero issues.  Assuming my memory is serving me correctly, both cards were run with a mild 250Mhz+ on the GPU cores. Just looking for others input on this matter, and I am up for any ideas! Attached is a comparison photo of the runs.

Yeah, Ivy Bridge is running PCI-E 3.0 which has less latency as well... As for your overclock, it could be a result in changing the PSU. 

No, they are not. The PCI-E lanes are on the CPU, not the motherboard. Also, overclocking causes degradation... So it could have all ready lost it's stability and you are just now starting to notice running through the stability tests.

Some high end P67 boards use a NF200 chip that does add more PCI-E lanes. Those extra lanes are PCI-E 2.0.

Well as an update to this performance impact, I decided to Update my P67 FTW to the latest bios, and re-install my 3570K and ramp it up to 4.2Ghz like before.  I actually came up with much worse performance scores then with the 2500k installed! So it lead's me to believe the PCI-3.0 wasn't the underlying performance difference. I may just Nuke & Pave & see how things come out. Perhaps the P67 FTW just ins't the performer like my Sabertooth P67 was.