Here is an article from our very own Tomshardware.com provening that the FX CPUs bottleneck frame rates more or less all the time, not just in the very CPU intensive games like BF3 at the highest settings.
http://www.tomshardware.com/reviews/core-i5-fx-6100-overclock-benchmark,3099-3.html
Some times the Bulldozer build lost to the Sandy Bridge build with half the budget. Admittedly the FX-6100 was a horrible choice for this comparison, an FX 8150 or at least an FX 8120 should have been tested but it does show the horrible per clock performance of Bulldozer none the less since the six core CPU is significantly slower than the quad core CPU with a lower clock frequency.
Increased frequency and core count did not save the FX lineup form the fact that the Bulldozer is a horrible attempt at making a CPU and it is still slower than One of Intel's cheaper quad core CPUs. This isn't even because of the modular approach of Bulldozer but it's performance is simply because AMD screwed it all up. The modular approach is, however, better geared towards servers and such highly threaded work and hurts at lightly threaded workloads. Regardless of this Bulldozer should be faster but there are several problems that could have been fixed.
One is stated by a former (high ranking) employee of AMD that tells us the automatic method of designing each transistor used to make the chips reduced performance by about 20% and increased power usage by about 20% when compared with how fast Bulldozer chips would be if they were hand-designed chips like pretty much all other CPUs made so far. I have to hope this is true and not a hoax because if AMD actually made this garbage called Bulldozer properly yet it fails so much then I see a bleak future where AMD has completely left the high performance/enthusiast market...
Another problem that has been recurring throughout some of AMD's designs is the CPU cache's low speeds compared to Intel's CPU cache. AMD increases the capacity of the cache on their CPUs in an attempt to bypass this problem but it doesn't seem to be working.
Take a hint from Intel, small but fast L1 and L2 caches with a high capacity L3 cache has better performance than larger but slower L1 and L2 caches with a high capacity L3 cache have. This has been tested and proven before and I have A PDF file that shows and slightly explains this phenomenon, although it is a little old.
There are even more problems but these two are probably the greatest factors that I can remember right now. If both of these were fixed then Bulldozer would be a great product that I could throw recommendations around for but no, AMD screwed it up.
I'm not even an expert but I can see these problems... Come on AMD this isn't funny.