One thing I neglected to mention was before I uninstalled the drivers I disabled the video card in device manager. So disable the card... uninstall the drivers then reboot. Enable the video card again then when I installed the drivers I only installed for the GPU and PhysX. 

I've read the same thing about using nvidia inspector but for some reason for me it made my system unstable. 

I would be careful modifying memory clock speeds.
 
As I interpret Manuel's reply, they clock down in those situations because it MIGHT be necessary to ensure 100% precise calculations. (Something which in games might not be a problem since games usually don't use heavy floating point calculations etc...)
 
The problem with Video memory is that you won't necessarily see when it encounters errors due to internal CRC correction.
 
In other words: Yes, you can adjust to "normal" speed in Maxwell Bios Editor etc. but you might not be aware in case there is computation errors. You might suffer performance when the memory "internally" has to correct itself. I figure there is a reason WHY Nvidia runs the memory clock lower, otherwise they wouldn't do this.

Where did you get the idea that "Cuda on the GTX 970 is a dead end"?
 
What the title means is that in CUDA apps, assuming the card stays in a lower power-state, the memory is running somewhat slower than, say, in Games. Nvidia officially explained this with that for computation, precision is more important than speed, therefore they're running the memory there slightly slower to prevent potential computation errors due to faster clocked memory.
 
While with a BIOS hack this could possibly be be "solved", it's a FAR CRY from saying that CUDA rendering is a "dead end" just because the memory is clocked slower. I figure that NV had a reason WHY they did this...so I wouldn't necessarily recommend to anyone to mod their bios because of this, especially since memory clocks do not have THAT large impact on performance. Of course it has, but TBH whether memory runs at 6000 or 7000 doesn't make such an enormous difference, at least not a significant one. This issue is way blown out of proportion, IMHO.

If it doesn't even utilize it at all (which I don't know at the moment) it's a driver issue. Nvidia did something with Cuda support in the latest drivers. But then an OLDER driver who supports CUDA should solve this? Sorry can't say too much about this but some Googling should shed some light on this.

I guess my first question to you is,

Have you actually tried running the computer with the GTX970 and your Premier Pro CS6?  What do the numbers look like compared to your old computer?
 
As far as update to Cuda by Nvidia.  I don't think there has been any changes to CUDA. There was a change to OpenCL from 1.1 to 1.2 but that won't affect CS6 as per their forums CS6 disables OpenCL.  So that is a mute point.
 
Is it possible that CS6 doesn't support the 970, yes it is as the last nvidia card I see on their support list is GTX 580. That was some time ago.
 
For anyone wondering about the change in driver 350.12 and how it affect crunching data. For those projects which use OpenCl, they will need to adjust their apps to compensate for the change (Seti as already done this with a new release) The only other advice to avoid this  issue is to stay with a driver below 350.12