Just food for thought,
Trinity at 32nm = 246mm2 1.3B transistors.
For simplicity, dual Module occupies half of the die.
Now just imagine what would happen if they would have a 22nm process next year.
1: They could have a 22nm Trinity at ~130mm2 with 30-50% less power
That would put them in the same die size as Core i3, same CPU performance but with 100% more iGPU performance and close the same power consumption.
2: They could have a Quad Module (8 cores SR) plus 384 GCN Shaders at ~190mm2
That would be a FX8350 + 15% more performance from SteamRoller cores + Trinitys iGPU but with GCN shaders and lower consumption (65watts ??)
3: They could have a Quad Module (8 cores SR) plus 512 GCN shaders at ~250mm2 and same TDP as Trinity.
That would have SteamRoller performance + 50% more iGPU performance than Trinity.
4: Dual Module (SteamRoller) + 512 GCN Shaders at ~160mm2.
The iGPU would get more space than the CPU cores. 15% more CPU performance + 50% more iGPU performance at 65W TDP.
I believe number 4 is the 28nm Kaveri at 100W TDP and close to ~220mm2 die size.
Just for reference, Intel 4C Ivy at 22nm = 160mm2 (GT2)
Your math is a little off.
When Intel moved from 32nm to 22nm, they went from 149mm² (2c Sandy) to 94mm² (2c Ivy) and from 216mm² (4c Sandy) to 160mm² (4c Ivy).
If AMD scales down similarly, Trinity would be 155-180mm², not 130mm², compared to the i3 which is 94mm².
Vishera would be 200-230mm² WITHOUT an iGPU, not a chance it'll get near 180mm² with 384 GCN cores, competing with 160mm² i7's with HD4k graphics (or i5's if you're comparing a cutdown version of the chip).
Interestingly, Trinity even at 22nm would be bigger than an Ivy Bridge i7, because it's already bigger than a Sandy Bridge i7.
I think Intel did very well considering their transistor and die size budget.