According to your
project list i would say, you have to many projects active...
Try to focus your hosts (are there really 11 hosts or did you forget to merge them) to only some projects. Your HD7970 is a good card but penalized by AMD with dev-manuals of lower quality. Developing for AMD seems to be more a trial and error thing. Further kicks in the back are OpenCL, available only on newer cards (HD5000 and up, HD4000 was and is beta), and discontinued support of the long used predecessor CAL. Only a few projects have released OpenCL apps and are able to bring the gpu power on the road. All other support Cuda because it is better documented and therefore more attractive for developers.
Your FX8350 is also known as bulldozer. There are many threads about it in the web. A BD performs good with multithreaded apps and is slower than a Phenom II if you use singlethreaded apps. You have only an octa-core cpu if you calculate integer values. Floating point operations are calculated by one execution unit per cpu-modul with two cores. FP-wise this means you own only a quad-core and the same is valid for AVX. AVX (256bit) can't be faster than SSE (128bit) because AVX needs both cores of a cpu-modul.
The short story is. Your cpu has a higher energy consumption in comparison to Sandy or especially Ivy Bridge and is slower. Only AMDs cpu feature FMA could make a difference but Intel will integrate FMA support in the coming Haswell. AMD is again the first in the integration of features and functions but wins only the second prize if you want to use it practically. On the other side nobody will develop extensions, that are only executable by one vendor if he is not the biggest player in this area...
@Stacie: Something with certainty and confirmation that you are a human and not an alien.