The bug was in "-=" operator in vector class.
you can issue the following command to run the application:
> SimpleRayTracer.exe [eyeX] [eyeY] [eyeZ] [atX] [atY] [atZ] [upX] [upY] [upZ] [width] [height]
on debug mode it takes 52" on intel Atom 330 1.6GHz and on release mode it takes 8" to complete the rendering.
when I implemented the reflection the rendering time goes from 22" to 52" on 3 bounce reflection.
