Would be nice, if we wouldn't need to care about the hardware reality and just by programming it with software we could double it - make out of 4 cores simply 8 cores calculating power.
If the hardware isn't there, it won't work - like having a 100 horsepower engine in your car, and by running a special control program you will have suddenly 200 hp. Using a car engine as comparison - such programs can have some effect on the effeciency of the engine running somehow better, but mostly helping to reduce fuel consumption or other important aspects of a car engine.
That doesn't work - like jismi explained it perfectly - it can only help to optimize the calculation and data handling process - which can give you some extra speed, but it won't be noticable in all cases of your work.

In a CPU with 4 cores you can say, that each core is a real single CPU, which can and could work separatly from each other - not depending on each other and by multi-core programming their power can be combined - like giving each core a piece of the cake while rendering an image for instance - just like sending parts of the job to other computers in a network/renderfarm.
Meaning - if 3 cores would be damaged and only one core would be left, this one core still could run the OS and applications, just slower (okay that's just hypotheticlly spoken - the CPU probably would be totally damaged then ...).

On a graphics card cores it's a different story - they can't do any work independently from each other. It's a special architecture design, that has to work together, so one single core can't work alone and can do everything each other can, because they are grouped, each group given certain tasks to perfom and only those tasks.
To be able to do their work, they need to be combined in groups, newly called SM - streaming multiprocessors, to be able to do their work. In the recent Nvidia GT200 series there are 8 of those cores combined to one SM - so all together there are 30 of those SMs - each SM doing something special (simply spoken - i'm not really exact in that, just to make it easier to understand - like one SM is calculating only shadows, the other is for reflections and so on ...)

Looking at the soon to be released new Nvidia graphic cards like the G300 with the new Fermi architecture, it got 512 Cuda cores - it's more than doubling the predecessor GT200 series (like GT240,G260,G280,G295 ...) in core and transistor numbers.
Here you will have 32 cores forming a SM - meaning the new Fermi technology is suppose to have 8x the combined power of the GT200 series ...

That's what i'm looking forward too. They have DirectX 11, too, fitting perfect to the new Windows 7 OS ...

Surely it would be nice, we could add the GPU on a graphic card to the rendering process, since it's doing barely nothing at all, while the CPU is rendering at full power. But like discribed above, it won't be that easy, to stear and handle the data stream to and from those SMs and cores, because they are usually used for different, specialized stuff relating to realtime 3D.

Photoshop CS4 started to use GPU calculation in a few effects / features such as realtime zoom and rotating and a few other things ...

But since they are different and use a different technical architecture while processing data, like FrankT mentioned, it must be more complicated to add them to a rendering process, otherwise it would have been done already since a longer time ... doesn't matter if we are talking about Vue, or other well known expensive highend graphic packages like C4D, 3dsMax, Houdini and whatever they are called ...