The dimensions used in the x, y, Width, and Height members of the D3DVIEWPORT8 type for a viewport define the location and dimensions of the viewport on the render-target surface. These values are in screen coordinates, relative to the upper-left corner of the surface.

Microsoft® Direct3D® uses the viewport location and dimensions to scale the vertices to fit a rendered scene into the appropriate location on the target surface. Internally, Direct3D inserts these values into a matrix that is applied to each vertex:

This matrix simply scales vertices according to the viewport dimensions and desired depth range and translates them to the appropriate location on the render-target surface. The matrix also flips the y-coordinate to reflect a screen origin at the top-left corner with y increasing downward. After this matrix is applied, vertices are still homogeneous—that is, they still exist as [x,y,z,w] vertices—and they must be converted to non-homogeneous coordinates before being sent to the rasterizer. This is performed by way of simple division.