I suspect this technique will need a lot of refinement before its actually competitive.
The main trouble with this technique, I suspect, is it doesn't get at the root of where power is really spent. Yes, anything costs power, but is that where all the juice is really going? A famous stat from the P4 days was, 40% of chip power was spent just driving the clocks. I'm sure it's better today, but how much? What about all the leakage in the SRAMs?
The main trouble with this technique, I suspect, is it doesn't get at the root of where power is really spent. Yes, anything costs power, but is that where all the juice is really going? A famous stat from the P4 days was, 40% of chip power was spent just driving the clocks. I'm sure it's better today, but how much? What about all the leakage in the SRAMs?