I read it only now..
Well powergating is the best hardware implementation ever!!
One thing.. since you are using the s4 octa.. are you quite sure kernel scheduler doesen't "send" tasks to 1 core when it's not necessary to permit a shut down of the correct transistors?
Because in theory, 3.10 scheduler can do it, but, don't know if it works with the 3.4 scheduler.. i'm meaning.. when i see the core load is 0, in theory that core is automatically shutted down by the powergating tech.. right?