In Muxed design both GPUs (the Radeon and the Intel) are connected to a hardware mux (switch), and the mux connects to the display. When you switch graphics you are changing which GPU connects to the display.
Muxless switching is the basic design of NVIDIA Optimus. What that means is that the NVIDIA GPU renders into memory, but the Intel GPU is used for outputting the final image. This is the key here - the Intel GPU is always active, it's just that the NVIDIA does the rendering work when it's selected. So the display is always connected to the Intel card which in turn is connected to the NVIDIA card.
The downsides of the muxless design are driver issues ,slightly lower performance (a couple of percent - not really noticeable), and
basically no support for the dedicated GPU under non-Windows operating systems (there are significant issues in Linux that make doing muxless designs harder, and the engineering resources are not there).
The upsides of the muxless design are much quicker and more transparent switching, better driver support in the long term (the mux-based design is laptop specific and rely on hacks that Microsoft doesn't really support), and lower cost.
[./quote]
However, something called "Bumblebee" for Linux has made it possible to switch Optimus GPUs in Linux. We just need someone to port it to OSX.
http://www.omgubuntu.co.uk/2011/05/bumb ... nux-users/
https://github.com/Bumblebee-Project/Bumblebee
Now looking at my specs, you'll might ask why I am interested in Optimus while my laptop has a AMD card. This is because AMD also decided to use "Muxless" switching in newer GPUs & my laptop has the same issues as the Optimus owners.