In immediate mode, the scene (complete object model of the rendering primitives) is retained in the memory space of the client, instead of the graphics library. This implies that in an immediate mode application the lists of graphical objects to be rendered are kept by the client and are not saved by the graphics library API. The application must re-issue all drawing commands required to describe the entire scene each time a new frame is required, regardless of actual changes. This method provides on the one hand a maximum of control and flexibility to the application program, but on the other hand it also generates continuous work load on the CPU.

Examples of immediate mode rendering systems include Direct2D,^[1] OpenGL^[3] and Quartz.^[4] There are some immediate mode GUIs that are particularly suitable when used in conjunction with immediate mode rendering systems.

Primitive vertex attribute data may be inserted frame by frame into a command buffer by a rendering API. This involves significant bandwidth and processor time (especially if the graphics processing unit is on a separate bus), but may be advantageous for data generated dynamically by the CPU. It is less common since the advent of increasingly versatile shaders, with which a graphics processing unit may generate increasingly complex effects without the need for CPU intervention.

Although drawing commands have to be re-issued for each new frame, modern systems using this method are generally able to avoid the unnecessary duplication of more memory-intensive display data by referring to that unchanging data (via indirection) (e.g. textures and vertex buffers) in the drawing commands.

• Immediate mode GUI
• Display list
• Retained mode, usually contrasted with immediate mode
• Scene graph