Also, does the x264 HD 5.0.1 benchmark use Microsoft's compiler? Cinebench uses ICC or OpenMP, and as the earlier LLVM/Open64/GCC tests show, compiler matters. (Note: I was looking for an ICC benchmark on Phoronix but couldn't find it.)
Extremetech's prime95 benchmark is very misleading. The core i5 can't work as hard because it's limited to 1/2 the number of workers as the fx-8350. I'm unfamilar with ET's site, is there an 8-thread comparison? This is important because at a transistor increase of 200M, it would help indicate the Ivy Bridge architecture efficiency.
Legitreview's prime95 page with a chart lists Battlefield 3, CPU load, and idle. The text above says it's supposed to be 3dMark11, which is still a black box as far as optimization is concerned.
I like the review. I can now see why early speculation said Bulldozer was going to catch up to SB -- and why AMD resigned from BAPCo. I imagine if AMD went under then you'd have the last benchmark site standing; no one else is capable of comparing as many benchmarks on POWER and ARM.