From the research I've conducted, most consumer software (ie: Sony, Adobe, and so on) don't really benefit too much from having computers with more than four physical cores (like the 4790k). The software is designed to utilise a certain amount of threads, and having more won't actually speed anything up; there's kind of nothing you can do about this, aside from switching software. Most consumer software I've seen doesn't really support more than 8 threads (my research could very well be wrong). If you're actually looking at what computer would perform better with which CPU, then shoot an email to Adobe. I'm willing to be Adobe CC will perform nearly identical with either the dual Xeon, or the 5960x, as they both are just not what the software is designed to utilise.
With that said,
in this article, it seems like CUDA processing is more natively supported by Adobe. This is an outdated article, but it does make the point that they've been supporting the CUDA technology natively for a bit longer than OpenCL; so I'd consider going the NVIDIA route for your GPU. Personally, I would prefer NVIDIA, as the AMD cards just don't seem to be supported as well with most things. They may work "fine" or "okay," but compared side-by-side, I'm willing to bet the NVIDIA cards would be notably smoother. Could be wrong, but that's what my research has led me to believe.
As far as your build is concerned, I don't see why you should consider going with an i7, if you're looking for maximum calculation power. You will have more brute force with dual Xeons, as you will have 24 threads at your command, whereas the i7 will only give you 16. Yes, you will benefit in some ways with a higher clock speed, but it also depends on what you're doing. In theory, if you were to match both the dual Xeon setup with the 5960x, in a video rendering test, the extra threads would help your Xeons get everything done faster. Just an example, though.