The Natural Convection Test Rig
The Natural Convection Test Rig-This page is dedicated to research into Pumpless, Fanless cooling of CPUs -using the Thermo-Siphon Principle.
The basic idea here is to use a high wattage resistor to provide an easily controlled heat source, which can be used to test different waterblocks, tubing and radiators. The goal here is to see what works in a CPU water-cooling system with no pump or fan.
The potential for noise reduction with a system like this is huge, as CPU cooling is often the main source of noise in a PC. The potential for overclocking though, is limited by comparison to a system using a pump and fan-assisted radiator, though it should be on a par with a reasonable air-cooled heatsink.
So far, I have tested:
I chose this waterblock because it had 1/2" internal diameter pipes, has a top to bottom flow path and internal fins too.
My hypothesis was that the commercial waterblock would be too restrictive to the waterflow to be good in a pumpless system. (I did however, hope that I would be wrong).
The first test was to see how much power could be dissipated with no water in the system. The power was increased to the point where the waterblock reached a temperature of 55 degrees Celsius. The temperature was measured using an industrial thermocouple with a tiny tip.
The ambient temperature during both tests was 21.5 degrees Celsius.
Next, the system was filled with water and the air pockets were bled out. The test was repeated.
| Test Info. | Power Dissipated |
|---|---|
| No Water in System | 4.5W |
| Radiator 30 deg. | 51.0W |
Having said that, the commercial waterblock did better than I had expected.
Watching small specs of dirt flow round the system showed how slowly the water was flowing. I calculate the flow rate was around 24 litres per hour. The fact that waterblock can dissipate so much heat with so little water flow is worth noting.
Commercial waterblocks are attractive because they can be readily bought, easily fitted and even come with a guarantee. They are brilliant when used with a pump but ultimately, their internal friction is rather too high for a pumpless system. If anyone wants to make their own pumpless watercooled PC, (like me, you will be in a minority :-) *and* you want to use a commercial waterblock, the best advice I can give is use a low wattage processor (consider undervolting it) and a waterblock with the biggest internal diameter pipes you can find. Also, make sure you use a waterblock where the water has a straight path to flow through or the thermo-siphon effect will not work properly.
Perhaps if some company makes a monster waterblock with larger diameter pipes, I will test one, but for now I will make a better custom one.
If the radiator is vertical, this improves the thermo-siphon effect which drives the water round, as the cooler (denser) water falls due to gravity, but the air flow is not encouraged to flow in a particular direction and the rising hot air envelops the radiator, keeping away cooler air to some extent.
If the radiator is horizontal however, this suits air flow, as cool air can enter on one side and pass directly up through the matrix and out of the other side, but the cool water is not biased into a particular flow direction and the waterblock design must be relied on to provide the flow direction.
It is of course possible to design a radiator whose air and water flow are in the same direction. Vehicle intercoolers are cheap though and do a reasonable job of dissipating the small heat output of a PC processor.
The best logical compromise of 45 degrees radiator angle might not be the best in practice. To see what the best angle is for my system and to get an idea of how much difference the angle makes anyway, I decided to do a series of tests.
In each test, the power was increased until the temperature of the waterblock reached 55 deg. C. If the temperature crept any higher, the power was reduced and then I waited again for the temperature to stabilize. (The tests took hours for this reason).
During the tests, the ambient temperature was 23.5 degrees Celsius (74 deg. F) The dissipation figures speak for themselves but it's interesting to note that the waterblock can only cope with 8.6W before the temperature rises to 55 deg. C.
I appreciate that the power resistor that supplies the heat cannot direct all of its heat into the waterblock, but as you see, the combined resistor and waterblock can only dissipate a small amount of heat before they reach 55deg. C. and since the waterblock has a much greater surface area than the resistor, it is therefore safe to assume that the power dissipated by the resistor represents a small amount of the total in the tests that have water in the system.
| Test Info. | Power Dissipated |
|---|---|
| No Water in System | 8.6W |
| Radiator 0 deg. (Horizontal) | 56.1W |
| Radiator 30 deg. | 67.0W |
| Radiator 45 deg. | 63.9W |
| Radiator 60 deg. | 56.1W |
| Radiator 90 deg. (Vertical) | 54.9W |
This setup works best when it is angled to favour airflow more than water flow. This suggests that the air is struggling to pass through the matrix. Vehicle intercoolers (like the one I am using as a radiator) are designed to dissipate kilowatts of heat, not a few tens of Watts. In order to maximise intercooler efficiency, the fins are closely spaced and even louvered to increase the surface area and promote turbulence. The resistance the air experiences as it flows through the matrix is not a problem in a moving vehicle, but it becomes the limiting factor when the radiator is used passively.
What with feeding 60 plus Watts of heat into the cooling rig, the thing that is getting hottest was, by a long way, my bench PSU. Only one thing for it, - get the lid off and lift it up so the air can get to it. Photo: shock hazard
Having finished the series of tests on radiator angles, I can now report that with this particular setup, the best angle (for THIS system) is 30 degrees from horizontal. This figure could be different when using other components of course and won't make any difference in systems using a pump... but it gives ME a point to work from in my experiments.
The case for using a vehicle intercooler to cool a computer is a curious one, but it comes down to cost -a radiator or intercooler from a car breakers yard should not cost too much. Passive radiators specifically for watercooled computers however are few and far between. I was recently delighted to see Innovatek have introduced a couple of models: the 'HTCS-Radiator' and the 'innovaKonvekt-O-Matic' Of particular interest is the design of the fins. Long widely spaced fins are best for natural air dissipation as they don't impede the rising air too much. Sadly, the design of the 'HTCS-Radiator' model does not lend itself to pumpless watercooling as a pump is needed to drive the water through it. The 'innovaKonvekt-O-Matic' model however, looks more suitable, though I have my doubts about the small diameter entry and exit holes the flow must pass through. This radiator is also designed to have water pumped through it but the simple vertical path that the water and airflow take make it look like it could be modified and used in a passive flow system.
If you are struggling to make sense of the Innovatek website (which is rich in product details but written in German) or you want to see some prices then visit 
www.KoolnQuiet.co.uk/ Personally, I quite like browsing round German sites -it doesn't take long to work out what the most common words mean and many of the words can be guessed from their similarity to words I know. And there is always the 'babel fish' ...
So what *is* the ideal radiator for a pumpless flow watercooled computer? -Simple answer is I don't yet know. Certainly, it will be mounted on the side or sides of a computer case, and certainly it will be made of specially designed extrusions, probably aluminium ones for ease of manufacture, cost and weight reasons. I suspect that the fins will need to be quite long and widely spaced. I would also speculate that the inlet and outlet pipes will need to be quite large in diameter. As I test more components, I should be able to supply the answers to most of these questions.
This is a photo of the high power resistor which I am using to apply heat to the cooling circuit. This will not perfectly simulate a computer CPU, but will help me know whether my experiment results are getting better or worse. Eventually, I will build passive cooling into my PC, using knowledge from these experiments.
General View of the 'Test Area'
