Presumably it means that software doesn’t have to write the same sort of layer offloading support. It’ll “just work” as if you had X GB of VRAM all along.

so, magic?