Unified memory
One memory pool shared by CPU and GPU, as on Apple Silicon and some AMD systems.
Unified memory lets a machine hold models far larger than any consumer discrete GPU, since the entire system memory is addressable by the GPU. The trade is bandwidth: even high-end unified systems sit below flagship discrete cards, and prompt processing is typically weaker.
Two details are routinely missed. Not all of the pool is allocatable to the GPU — Apple reserves roughly a quarter by default, and some AMD systems cap the GPU share explicitly. And the same marketing name can ship at different bus widths, so an identical chip name can differ by a third in bandwidth depending on the memory configuration ordered.