The description of matBones (which is, for me, one of the features I've been looking forward to most) says "With only a few bones and no weights, the difference is less noticable." -- I was under the impression that the amount of vertices attached to animated bones was by far more of an issue than the number of bones.
I think any decent gaming graphics card will be vastly more powerful regardless.
Since gpu animation will run on the vertex shader, and the vertex shader is the most underutilized part of rendering, there is plenty of headroom for the gpu.
I forgot to mention that animation can also be accomplished within a fixed function effect on hardware previous to vertex shaders.
In this
old article they compare a Radeon 9700 being 5 times faster than a pentium 4 at 3.0 ghz. That card did not have universal shaders, so there was less vertex processors than pixel processors, and it STILL got 5 times performance.
Now we have dynamic allocation to vertex and pixel shaders, and shader performance expanded vastly since then.
