In Sebbi's response he seems to have forgotten about the 16 shadow casting lights that generate 64MB of data time 2 (128MB) ... Since the EDRAM is already full with the frame buffer data sets that's becoming quite limiting.
Let's say for a minute that this would equalize the bandwidth, then you hit the real wall, the GPU, it has only so many compute/shader units, limited ROPS and ACEs, all of which are significantly lacking on the XB1...
TL:DR; MS was right when they said their architecture was balanced, the GPU they picked does not need more memory bandwidth, it would probably go unused most of the time.







