Interesting that the system switches between two distinct devices at different power/performance points. There is probably some redundancy here, in that the higher power/performance (discrete) GPU likely includes all the elements of the simplified (integrated Intel) GPU. So the discrete GPU could be designed in a way that dials down the power and performance to match the discrete CPU. I know Intel and AMD CPUs include voltage/frequency/power-gating dials for selecting a power/performance point, but I don't know if GPU engines make heavy use of these features.
The simplified GPU being integrated with the CPU probably saves some power from not having to go off-chip to talk to each other. This power savings might be what justifies having two distinct devices.
Most of us who bought the previous models were under the distinct impression that a future OS upgrade would let us switch without logging in/out, and that we could take advantage of the unused dedicated GPU for other stuff as well.
I don't personally recall seeing anything to support this. Can you expand on who/what gave you that impression, in case I just missed it?
The article suggests that this is both a hardware and software solution, so that it would be impossible to add to older hardware not specifically designed with these new techniques in mind:
"...because Apple is unique among PC vendors in that it controls both the hardware and software."
"NVIDIA's Optimus is another solution to this problem. It uses a combination of hardware and software to enable automatic switching between integrated graphics and a discrete GPU."
The simplified GPU being integrated with the CPU probably saves some power from not having to go off-chip to talk to each other. This power savings might be what justifies having two distinct devices.