If its faster than quadcore Llano clock for clock with two modules, im surprised. With change to vliw4 arch it will get more efficiency from same area but memory bw increase is where most of gpu potential is. 30% for GPU should be easy.