In one of its presentations at Meltdown, Nvidia made the following two statements:

The advent of programmable vertex shaders and pixel shaders in hardware is the most fundamental change in graphics for a long time! But, many developers do not use them in their games, or only use them to map simple multipass shaders into a single pass

The presenter goes on to encourage the use of VS and PS for shadow volume generation from stencil shadows, and order independent transparency. I've tried to explain the two techniques, briefly, in the following diagram captions.

Suffice to say that it is obvious that there is still a need for evangelizing, even by Nvidia, to get the game development community to deliver noticeable advantages on next generation GPUs. It kind of puts everything we've talked about here in perspective.

On the other hand, the elegance of the effects created using the power of something like the GeForce3 VS and PS doesn't necessarily translate into real world benefits. Great, realistic shadows are good at one level, but in a fast moving action game, how good does the shadow have to be? How realistic? Are you watching me, or watching my shadow?

This is beautiful. If you can get this kind of shadowing in a game then, you have to be impressed. To get to this level of quality, you need to get the VS to extrude the shadow volumes away from the light (little white point in top right hand corner) to create shadow volumes. The quick and dirty method would only do an approximate silhouette so, it wouldn't look so neat and clean for complex objects. Fine on spheres and cubes, but not much else.