I'm kinda confused by the "number of vertex program outputs" limit: document of DX10 says DX10 card has 32 interpolated registers (32 float4 output) while my card 9600GT reports its MAX_VARYING_FLOATS is 60 which means only 15 vec4 outputs are available. Why does NVidia make us to use a portion of its capacity in OpenGL while DX10 can access all of its registers? dex

I think it's 32 only for DX 10.1, for DX 10.0 only 16 was mandatory. Also remember that there's usually more registers available between geometry and fragment stages (MAX_GEOMETRY_VARYING_COMPONENTS_EXT returns 124 that is 31 x vec4 on my 8800GT)

thanks