A cube map is an image consisting of 6 layers, and each layer is drawn onto one face of an invisible cube. When the reflective surface is in view, it reflects the faces of the cube:
On the left you can see a list of layers, and on the right is the currently active layer. The layer names are used to define the face that layer will be drawn on: positive x is the north face of the cube, and negative x is the south face. The Y faces are the top and bottom and the Z faces are east/west.
So, cube maps prevent excessive load on graphics cards by simplifying the reflections. Reflecting a static texture is one thing, but if the reflections were based on the actual track, the load on the graphics card would probably be more than twice as high. E.g. perspective would have to be taken into account, as would texture transparency and the motion of other objects (bear in mind that you could have 30 cars driving past, maybe with headlights/brake-lights shining), and distortions due to the angle of each face of the reflective surface.
Programs such as 3dsMax and 3DSimed allow cube map files to be set via the material properties. When choosing the Shader level, you should see a number of settings that involve cube maps:
The best choice depends on whether or not you also want the material have bump maps, specular maps etc.
