i built an all electronic software controlled cooling system. gone is the old belt-driving cast-iron monstrosity and the thermostat. in it's place, physically, is a Davies Craig EWP80 electric pump, and a small Bosch secondary circulation pump. the radiator is a low cost generic "Ford type" aluminum pump from Speedway. there is no thermostat. temperature control is done by computer code in the chassis computer reading a pair of sensors (head, radiator outlet) many times per second. the cooling fan is electric. the sole remaining belt-driven accessory is the alternator.
at absolute maximum cooling, which occurs rarely, power consumption is about 300 watts, or a third of a horsepower. typical power consumption is less than 50 watts, about 3 amperes. a system for a large V8 would be maybe twice this. the EWP80 is the smallest pump of the three that Davies Craig makes.
i made the block hose adapter from a nasty old leaky water pump. drove out the shaft, chopped the nose and neck off, milled it flat, screwed on a cover, and an internal deflector that mimics the original impeller to avoid a cold spot on the #1 cylinder.
the factory steel thermostat housings are all rotted, and inconvenient to use with the aluminum radiator. i made a not completely satisfying adapter from an aluminum slab that angles the plastic Davies Craig 90-degree neck down to point at the radiator top inlet. though it's lasted a couple years i want to replace it with an all-metal unit.
this is the Davies Craig EWP80 pump. it's small and light and so far quite reliable. it more or less hangs on its hoses, though i've tied it down with silicone foam to prevent it vibrating. the Bosch circulating pump is tiny (it's often called a "heater booster pump") and does hang on it's hoses, though i siliconed a pad to it's backside and restrained it slightly with a cable tie to keep it from bouncing around.
the radiator took some ingenuity to mount. rather than have someone weld brackets to the thin aluminum tanks (which would then likely stress fracture the tanks in a real life street car) i made a sheet and tube saddle for it. the saddle bolts to the stock mount flanges, is lined with high-temperature silicone foam (aircraft surplus) and the radiator fits inside it snugly. as a positive side effect this electrically isolates the aluminum radiator and eliminates electrochemical corrosion due to the dissimilar metals thing (iron block, aluminum radiator, hot conducting coolant).
since this whole thing is so grossly overkill, and it is unlikely i will ever spend time in actual freezing weather, i'm running distilled water with Redline "water wetter". i have no need to raise the boiling point beyond 212F.
to save the front of the radiator from bugs and gravel, and the rear from my fingers and dropped wrenches or whatever, i put honeycomb nomex over the radiator front and rear. it blocks airflow slightly, but doesn't seem to matter.
these cheap aluminum radiators don't come with drains. i added one -- plus a strainer to protect the close-tolerance plastic impeller in the pump -- to an adapter in the lower hose. i needed to fit the rad's large (1.75") outlet to the pump's small (1.25") fitting and this thing was perfect. it was also the place to put the radiator-outlet temperature sensor, necessary for the cooling calculations. the temperature sensor in the adapter needs a ground wire, as electrically the thing is floating.
you can see the tiny Bosch circulation pump in these next two photos. i drilled and tapped the thermostat pod in the head for 3/8" NPT (later heads come so equipped) and the 1963 water pump has 3/8" NPT for the heater connection. the circ pump pulls from the head (warm) and pushes into the block (cool). the main pump pulls from the bottom of the radiator (radiator fills from the top, warm cooland from the head) and pushes into the block, same as the factory pump.
so it turns out that back in 1965 AMC solved the cylinder head thermal shock problem mentioned elsewhere by installing a "three legged" water pump, the top of which sucks water from the head side of the thermostat, and outputting to the block, even when the thermostat is closed. i imagine that this actually detracts from the absolutely cooling ability of the system, but seeing how this was a last-model-year fix, with the new-in-1964 new six a total success, this engine was fading away anyways. (just bought this junk engine from craigslist; i call it my "$200 valve cover" as it came with a chrome valve cover -- which turned out to be warped.)
needless to say this goes well beyond solving the cylinder head cooling design flaw. the circulation pump operates once there is detectable heat in the head; it then begins to circulate coolant between head and block only. this both distributes heat between head and block and ensures that the temperature sensors correctly indicate instantaneous coolant temperature, rather than relying on internal convection. circulation pump control is separate frmo the main coolant pump; it's speed is inversely proportional to the coolant pump (eg. when the coolant pump runs slow, circulation is fast, and vice versa).
the main coolant pump speed is calculated from the head and radiator temperature measurements, many times per second. pump speed necessarily anticipates temperature rise in the head in order to hold head temperature constant.
because the pump is electrical the computer can cool the engine down after it is shut off. when necessary the computer runs the pump and fan for up to three minutes after shutoff to bring the coolant temperature a few degrees below the setpoint to prevent underhood heat-soak.