The only advantageous ram move MS can realistically? do at this point is to increase the ESRAM but if that increases transistor count (I dunno) then it's an unrealistic move. Upping DDR3 from 8 to 16GB would be absolutely pointless.
Doesn't 32mb ESRAM already has a high amount of transistor count like half a 680GTX.
Doubling it would be hard i think. I would say kill the 8gb ddr3 and also go with gddr5.
Take the loss at launch and wait 2 ~ 3 years to replace it with 3d stacked memory down the line this will only get cheaper. And microsoft can probably recover the cost from the early adapters because they will pay longer to use xbox live gold.
Or just suck it up and release midway 2014 or early 2014.
We only start seeing the true power of a system in the 2nd or 3rd wave of games. Doing this wouldn't really change that (if they concealed things).
Im not sure we dont have the transition of going from single core system to multi core systems.
That transition was hard as fuck you need to think different, parallel programming is a fuck to debug because its not deterministic anymore with shit like race conditions.
Gpgpu is easy most devs could play with it for like 5 years already and SPU on ps3 is almost the same as programming for a compute unit.