AMD, nVidia and SM 3.0 - vertex texture fetch

12gauge

I am buying a new computer on which I want to work a bit with vertex and fragment shaders. This includes some calculations which I would like to do on the gpu (wave calculation for example). I want to do all of this under OpenGL with the GLSL. I have read the Ati cards do not support vertex texture fetch in the shader model 3.0 but instead proved another technique calles R2VB (render to vertex buffer?).
Does this even apply to my situation when I am using opengl? If yes, which technique is more used aka. which card would be better suited, a GF7900GS or a Radeon 1950XT? Gaming is secondary, but of course I would like to use the card for that too.

Clipse

For starters, nVidia supports* R2VB (yes, it does stand for Render to Vertex Buffer), so you won't lose anything by going with them. Furthermore, VTF allows you to do pretty much anything you would want to do with R2VB, and usually is simpler. The only advantage ATI has in this case is that the X1950XT is (I think) a faster card. Also, if your primary purpose for the card is scientific computing, you may want to look into ATI's CTM ("Close to the Metal"), which the 1950XT supports. nVidia's equivalent, CUDA, is not supported by anything earlier than the Geforce 8 series.

* - See the section on "Render to Vertex Array"