First, I disagree. The benchmarks show that the 543MP3 is fast enough if the memory design is altered to be more like the A6X.
Second, the A6+ I listed has a 32nm A6+ with a SGX 554MP2 which is faster than the current 543MP3 in the A6 or the 543MP4 in the A5X.
We go round and round but frankly you never accepted what is plain truth...the GPU in the A6 or half the GPU in the A6X is fast enough with a lot less silicon than is required in the A6X. The restructuring of the memory pipeline is already done for the A6X and packaging it with either half the cores (554MP2) or the existing GPU (543MP3) will work.
There was NOT paragraph after paragraph...there was one comment followed by a bunch of benchmarks and comments that showed the 543MP3 in the A6 was within a few percentage points of performance with the 543MP4 in the A5X. The one objection of memory bandwidth is a solved problem.
Moving the half the cores greatly reduces the footprint and power consumption and the iPad mini currently has a lot of headroom in terms of battery life.
I don't know how mine and Anand's comments on it can make it any more clear so good luck with Apple shoehorning a 2048x1536 into the iPad mini and using an unmodified A6 to do the job.