hmm... your CPU lists as 3.4GHz stock. Depending on how you are overclocking, you may not be getting the proper benefits you would otherwise. By it's current design (again.. DX9 is emulated in all versions of Windows past XP), the game will always be limited to a large degree by the CPU--although as always with these things, there is a plateau effect for performance gains where once you reach a certain threshold you move into the realm of diminishing returns. For me, that threshold seemed to be right around 4Ghz on the CPU for my C2D system. I do however continue to benefit from increasing memory performance because that system is still running on DDR2, so I have not met the threshold for diminishing returns for that subsystem--even when I've tweaked it to 1100Mhz and pushing right at 9GB/s. Note also the the CPU is also bound to Sound now as well, and it is geting harder to find dedicated network cards as well. The CPU is left managing a LOT more stuff now than it used to.
So, on that PC I can actually still see some slight scaling when ramping up the GPU memory (enhances fill rate), but ramping up the core speed provides minimal benefit. That is because of all the pre-processing that takes place within the game engine and the DX API's thunking layer, BEFORE it gets passed on to the GPU for post-processing. All of that early treatment is CPU bound, but you may have reached a point of diminishing returns as I had around the 4GHz mark. I can take mine to 4.3GHz in winter, but had to notch it back to 3.96 in the dead heat of summer. The frame rates were not impacted by any noticeable degree, so I just leave it there now since it allows me to strap the memory differently to get more transfer speed on that front.
It also explains why my quad-core 2.4GHz laptop with a slightly less robust GPU can match and at times slightly out-match my PC with a ~4GHz Dual Core with a stronger GPU. Though the laptop has a slower clocked CPU and is strapped to 100Mhz (PC uses 420Mhz), it has double the cores that enables it to off-set the lost workload (and is actually slightly higher--WEI CPU .2 higher on the laptop than the PC). That combines with the faster memory (DDR3-1600 (133 strap), vs DDR2-1100(220 strap), for .8 difference in indexing) to make things run much faster up front before it hands off to the weaker GPU (laptop scores 7.6 vs 8.1 on the PC)--but the laptop can sustain better FPS overall because it has slightly more punch on the CPU/MEM subsystems where the bulk of the crunching takes place.
Again, since you are clocking that CPU past 4GHz in both scenarios, the speeds you are running the CPU may be past the point that higher speeds gives a perceived benefit from raw CPU power. Depending on how you've configured it, you may actually be shooting yourself in the foot and you still have a bottleneck elsewhere (like strapping solely to 100Mhz or something--may not have options to change it, don't know), which could be restricting the potential (should also note my i7 at 2.4 scores nearly the same as your i5 at 3.4 according to Passmark). your issues could very well be more impacted by a fill rate problem and not processing power, and all that extra horsepower is simply going to waste. You also could be creating a high error rate as well by pushing components to their limit. I have seen systems run worse at a max theorized overclock than when you back them off just a tad from those max settings, because you are stressing the boundaries of the signal---you need a clean waveform for things to run properly.
Simply put... there could be something off-kilter elsewhere in your system causing issues.