Why does that power have to be used to generate more data than the GPU can handle? Think "quality" of data as much as quantity. A simple example could be trading off the number of particles in a system generated by Cell against the sophistication of the physical modelling you're applying to it. I actually think it would be a pity in many instances to simply use Cell to generate a lot of "dumb" data to be rendered.
That said, one of the first things nVidia mentioned about RSX was how they've never had a CPU that could feed a chip like it before, so that suggests that was one area of focus for them, how to handle that. But I don't think that need involve Visualiser patents and the like