The DirectDrawSurface object represents a surface that usually resides in the display memory, but can exist in system memory if display memory is exhausted or if it is explicitly requested.
Use the IDirectDraw2::CreateSurface method to create one surface or to simultaneously create multiple surfaces (a complex surface). When calling CreateSurface, you specify the dimensions of the surface, whether it is a single surface or a complex surface, and the pixel format (if the surface won't be using an indexed palette). All these characteristics are contained in a DDSURFACEDESC structure, whose address you send with the call. If the hardware can't support the requested capabilities or if it previously allocated those resources to another DirectDrawSurface object, the call will fail.
Creating single surfaces or multiple surfaces is a simple matter that requires only a few lines of code. There are four main scenarios for creating surfaces. Each scenario requires a little more preparation than the one before it, but none are difficult. The following four scenarios are discussed:
By default, DirectDraw attempts to create a surface in local video memory. If there isn't enough local video memory available to hold the surface, DirectDraw will try to use non-local video memory (on some AGP-equipped systems), and fall back on system memory if all other types of memory are unavailable. You can explicitly request that a surface be created in a certain type of memory by including the appropriate flags in the associated DDSCAPS structure when calling CreateSurface.