Up until now computer designs that have been theoretically known have been less energy efficient than the human brain. We could project supercomputers that could reach an exaFLOP and perhaps a zettaFLOP but the designs typically need megawatts to gigawatts of power.
ExaFLOP (10^18 operations per second)
Onchip photonics at low power a key aspect of taking something like our current computer hardware to zettaFLOP (10^21 operations per second) levels
Some common estimates of the computing power of the human brain are 10^13 to 10^16 operations per second 10^19 operations per second is a likely upper bound to simulate a human mind. The human brain needs about 20 watts of power.
A mechanical molecular computer designed of the Merkle-Freitas et al design has the potential to provide 10^21 Operations/Watt, over 10^11 times more efficient than conventional “green” supercomputers, which currently provide about 7 GFLOPS/Watt. Only two types of parts are required: Links, and rotary joints. Links are simply stiff, beam-like structures. Rotary joints are joints that allow rotational movement in a single plane.
It would be 2000 times more energy computation efficient than the upper bound estimate for the human brain.
A molecular model of a diamond-based lock, top view. Hydrogens are white, Carbons are green
A molecular model of a diamond-based lock, ¾ view