Just FYI, the TESLA models are not for Video or 3D, they are compute engine models meant to do large number crunching, and memory is vital to that, more vital than even large data sets for 2D and 3D which benefit more from compression.
The ATi Equivalents are;
Quadro = FireGL, FirePro and FireMV (there are sub models for Quadro like the NVS and they match up depending on features. They are 2D/3D/Multi-monitor cards.
Tesla = FireStream : they are GPGPU cards meant primarily to be used a co-processors or massively parrallel clusters.
Geforce = Radeon : these are the gaming lines
nforce is a motherboard chipset and driver package
Cuda is a programming platform for making the GPUs run non-graphical code, the ATi equivalents are CAL & Brook+ with CTM being a little more difficult and to the baseline where you essentially just code for the features/options not translate. OpenCL and DX11's compute shader support will also take part in this area soon and take some of their business as well as open up more to smaller players like S3.