"The group's goal is to create an architecture that melds graphics accelerators and application processors together into the same virtual memory address spaces so software can easily use GPUs to speed up tasks. It should also drive down power usage by simplifying the design of systems with fewer buses and controllers."