HappySqurriel said:
With that said, Nintendo is in a fairly unique position heading into this generation because they can freely choose whatever architecture they want. The reason for this is that they can maintain backwards compatibility with the Wii through software emulation because they will have the necessary processing power to pull it off (regardless of the architecture they choose). On top of that, most architectures could be made to suit the needs of a console manufacturer. You could take something which seemed underpowered for a game console (ARM processor or Intel Atom) and increase the number of cores, expand the instruction set, and/or over-clock it to make it perform at the level you desired; or you could take something that seemed to run too hot (Power 7) and scale it down by reducing the number of cores and/or under-clocking it to suit your needs. |
Well... Intel is only slated to start production of 22nm parts at the end of this year, it is unlikely that a 22nm console will be practical only a year after that given the increased difficulty in transitioning to new process nodes if you consider Global Foundries 32nm problems and TSMC's 40nm problems. 28nm is the safe option for releasing a new console and both GF and TSMC will have good 28nm processes by the time the console goes into production. Even better they may be significantly cheaper than the 22nm process at the time of the NES 6 release. It doesn't always follow that transitioning to a low node will reduce the cost per transistor, it may actually increase costs.
Tease.