Need help for test on CS6 Premiere Mercury Render Engine

[Jack Colquitt]

I try to understand what is being computed by graphic card cores, and what is being computed by processor cores.
Logically, the Mercury Engine (the CUDA cores of the graphic card) is used when we put a R3D on timeline,
and we add filters, resize, or color correction on it... but it's not relly clear if a "raw" R3D without filter
is read directly on the timeline by processor or graphic card.

So can someone put a R3D (without any correction or filters) on timeline, and read it, and next disable the Mercury Engine
and retry to read it at same resolution (or another R3D footage to be sure the first test file is not directly read from the RAM).
And if it's became slower, it logically mean that Premiere use mainly graphic card for reading timeline, for "RAW" or corrected footage.

The purpose of this test is to know what is more important on a R3D workstation, processor or graphic card.
Of course, both are important, and everything is used for preview....
processor, graphics card, RAM, SSD...
And we know it's possible to read full quality 4K files from Premiere timeline,
since the Olivier Madar test with a "simple" Clevo laptop i7-3920XM with a GTX670M.

http://www.reduser.net/forum/showth...tebook-on-Mac-i7-versus-PC-with-Premiere-CS6&

But I try to define clearly if it possible to buy a less powerfull processor (100 or 200 mHz less)
without losing ability to read full quality 4K/5K R3D on timeline, or if it's better to use a less powerfull graphic card.

In exemple i7-3920XM 2,9gHz is $1000 vs. i7-3840QM 2,8gHz at $500,
and the same between a $1500 GTX680M vs. a $1000 GTX675M.
And even if you put aside the hardware price, it's not really good for overall
computer temperature to use a hot 55W processor with a huge 100W graphic card !

Thanks !