It is unlikely that the current implementation of Nvidia's dual-GPU setup on the Macbook Pro will ever support concurrent use of both GPUs. Here's why:
Nvidia's "Hybrid SLI" is really two technologies:
1) HybridPower
HybridPower unleashes graphics performance when needed and switches to quiet, low-power quiet operation for everyday computing. Dial up performance for demanding 3D games and applications; downshift to the mGPU to reduce noise and extended battery life for everyday computing tasks like browsing the Web, word processing, or watching High Definition videos.
2) GeForce Boost
GeForce Boost turbocharges the performance of NVIDIA discrete GPUs when combined with NVIDIA motherboard GPUs. Plug any NVIDIA Hybrid SLI-enabled GPU into any NVIDIA Hybrid SLI-enabled motherboard to enjoy additive performance and more for your money.
The Macbook Pro uses number 1. What this does is switch between the integrated graphics and the discrete GPU when the operating system demands it. Right now, that's only on login/logout. In the future, this could potentially be more flexible.
The reason for this is that the Quartz Compositor (the software that allows GPU accelerated graphics) runs as a child of the loginwindow process. Although it could be uncoupled, right now the compositor is set up when the loginwindow creates an instance for a user at login time. This is pretty low-level functionality and so if that behavior changes, it's most likely to do so in Snow Leopard.
Technology number 2 is GeForce Boost. This is only available when the discrete graphic chip is a 9200, 9300, or 9500 chip. It is NOT available, mac or windows, for the 9600 and higher series. The most likely reason for this the nature of SLI. SLI means Scan Line Interleaving. With 2 GPUs, one GPU draws the odd number rows of pixels and the other draws the even rows. This generally works best when both of your GPUs are about the same speed. Otherwise, the faster GPU is just waiting for the slower one to finish drawing. This is why only the slower discrete GPUs are supported for GeForce Boost. It's because the 9200 series is in the same speed range as the on-board 9400.
It's likely that the discrete GPU in the MacBook Pro is faster on its own than a 9200 would be interleaved with the 9400. Essentially, GeForce Boost uses the following formula:
[Slowest GPU] x2 = peak performance
If the discrete GPU is 2x the speed of the slowest GPU (9400 integrated) then you get no gains from GeForce Boost. It's possible that you could do SLI in a [3 to 2] or [2 to 1] ratio but I'm not aware of any SLI systems on any platform that currently are capable of apportioning SLI rendering based on relative GPU speed. It always requires a matched or near-matched GPU pair.
Finally, you get a benefit from using just the discrete GPU by itself and turning off the 9400. Because the 9400 uses main memory instead of dedicated DDR, when it's active you have the GPU traffic going over the main memory bus. When you disable the 9400 in favor of the separate GPU, you free up some memory bandwidth (even if the RAM's still allocated to the GPU the bus traffic isn't active) and thus the CPU has more resources available. In a gaming or rendering situation, this is probably preferable to a GeForce Boost scenario if you main GPU is fast enough.
Another thing to consider is that Apple is way behind the Windows market in implementing GPU features. For instance, on my MacBook Pro, my CPU load is much higher playing an H.264 video in OSX than it is in Windows. Why? Most modern GPUs offer hardware acceleration of video playback. OSX Leopard currently does not use this functionality to its full extent and still renders the movie in software (but does do window display and compositing in hardware).
Hopefully this helps explain the MBP GPU setup a bit more.
