Has anyone tried a controller for cooling that is dependent on processor current consumption? Temperature measurements lag, but the current used by the processor is instantaneous and directly converted to heat (P = VI). In theory, it should be possible to reduce temperature spikes.