I've got a problem using DirectX3D. I've copied an example from the help and transformed it a little to work in c++builder. And when running the program I occassionaly get an Invalid Floating point operation error or a divide by zero error. The time it happens depends on the type of object I load in my 3d space, and the position of the objects. With a cube it doesn't happen, with a sphere it most of the time does.
I had exactly the same problem when I tried this with OpenGL.
With OpenGL I could see (by debugging) that the error occurs inside the OpenGL a function call (for rendering I think - can't remember).
Any suggestions would be appreciated,