The Lucid Chip sits on the motherboard to manage multiple GPUs as if they were a single resource. Alternatively it sits on a graphics card or daughter board on the PCIe bus and does the same thing.
So instead of the CPU/MCP seeing the lowest specced card (or only one with SLi) it sends data to the Lucid chip which passes it on to the GPUs according to their capabilities. Basically if you add a 4850 to 3870 at present the 4850 will be constrained by the R670. With Lucid managing data flow (perhaps adding latency) you would get performance scaling according to the capability of the card.
Basically it makes heterogeneous multiple GPUs work together in a way that ATi has been attempting and nVidia has been avoiding.
Bookmarks