The Fountain Cooler

Using an ornamental water fountain for CPU cooling. — Bruno M. Pancorbo and Paula K. Knabb.

Inspired by the number of water cooling systems and especially by the evaporative pipe type cooling systems, I decided it was time to set up my own rig. After researching the different types of cooling mechanisms (large heat sinks, radiators, water towers, peltiers and other combinations), I decided to stick with a simple water block plus water tower system.

Originally I built a pipe type water tower, and although I was impressed with its ability to cool the water a couple of degrees below room temperature, I was not very impressed with its water capacity. It was a rather small unit and the water evaporation rate was such that I would have had to “water” the computer every other day to keep enough water in the reservoir.

At that point I could have elected to build a larger version or to somehow enlarge the water reservoir to compensate for the high water evaporation rate, but I decided to try something slightly different the second time around. In this paper, I present a fountain water cooling system that is effective, very quiet and requires very little maintenance.

Materials

• Duron 650 @ 1000 MHz
• Plastic fountain (purchased for home decoration purposes)
• Water pump (that came with fountain, no name but quiet)
• Danger Dan Maze-2 copper block (~ $40)
• 60 mm 12 V computer fan
• Telko ® door sound alarm system (from Home Depot ~ $6)
• Dura Sept ® evaporative humidifier bacteriostatic treatment (from Home Depot ~ $4)
• 40 L trash can AKA water reservoir (~ $5)
  -Tubing, brackets (~ $5)

Lying around the house was a fountain

that my girlfriend got from QVC several months ago. The fountain is about 52″ (1.3 m) in height and 6 3/4″ (17 cm) wide. Its water reservoir capacity is about 10L, but due to the pump height, its effective water capacity is close to 7L instead. Since the fountain has been collecting dust I decided I would put it to a good use instead.

The first thing I did was to test the ability of the pump that came with the fountain to move water through the added tubing and water block. A quick experiment determined that the pump delivered about 1.5L/min (20 gal/hr); to my surprise, it was very quiet too. The next issue was to determine if the pump + fountain could actually keep up with a CPU’s heat.

For this test I used a K6-2 at 350 MHz. While these CPUs were very hot in their time, they are really nothing compared with the new Athlons, so this test was just to give me a rough estimate of the system’s ability to cool.

To make the story short, the K6-2 was cooled without any problems. With a heatsink + fan, the computer ran at 40 C at full load compared to 26 C using the water block (room temp @ 24 C.) Next, it was time for the Duron 650 @ 1000 MHz. I used this CPU because it is very stable at 1000 MHz with standard cooling (aluminum Global Win heatsink + fan). Table 1 shows the results.

Table 1. Temperature reading in degrees C for different setups of the fountain system. RT = Room temperature , WT = Water Temperature, and CPU temperature. Temperature measurements were taken with dual expanded range thermometer (-40 to 1200 range) with its probe about 1 mm from the CPU core between the waterblock and the CPU ceramic. All of the readings were taken with the case cover off. Idle = Win 2000, nothing running; Full = Prime 95.

The fountain is able to keep up with the Duron much better than the heatsink + fan combo, as expected for a water cooling system.

The water temperature in the fan-less system increases until equilibrium is reached 4 C and 6 C above RT for idle and full load, respectively. This shows that the fountain is able to cool the water without the need of a fan, although it is not dissipating heat fast enough to keep the water at room temperature. In contrast, both of the fan systems are able to keep the water temperature closer to RT. In fact, the 140 mm fan system goes as low as 3 degrees below RT at idle.

Another observation is the ability of all the fountain systems to keep the CPU temperature to about 9 C above water temperature (not RT). This makes sense because the difference between CPU temperature and water temperature should be determined by the type of heat block and the flow rate of the water, and since we are not changing any of those variables, the delta remains the same. To lower the delta between CPU temperature and water temperature we will need to either replace the heat block or increase the water flow rate (or both.)
{mospagebreak}

Bruno M. Pancorbo and Paula K. Knabb

As expected, the setup with the larger fan (140 mm) is clearly the better performer when compared to the 60 mm, although the difference is very small when the systems are under full load (30 and 31 degrees, respectively). The efficiency of the different setups is shown in Table 2:

Table 2. Cooling efficiency represented as C/W for the heatsink + fan combo and the three fountain setups.
*C/W values were calculated using RADIATE.
** This number was determined using numbers from table 1. Since those numbers were taken with the case cover off, it is very likely that the actual temperature of the enclosed system would be higher, thus lowering the efficiency even more for the heatsink + fan system but not for the fountain systems.

Here we can clearly see the difference between a standard heatsink + fan combo and a watercooling system. At 0.49 C/W, the heatsink + fan combo is very inefficient compared to 0.23, 0.16 and 0.12 C/W for the fountain setups at full load. Although there are many high end copper heatsinks that give much lower C/W values, most of them require a high speed (loud) fan for optimum performance, making them unsuitable for this particular project.

Based on the data from Table 2 alone, it appears that the fountain setup with the 140 mm fan is the way to go. Unfortunately, while this setup gives the best cooling, it has some disadvantages compared to the 60 mm setup. First of all was the noise.

All other things being equal, the larger fan produces more noise than the smaller fan. Because the 60 mm fan I am using is extremely quiet, the larger fan just can’t win in the noise department. The second reason, and far more important in this case, was the water evaporation rate of the 140 mm fan setups. As can be seen in Table 3, there is a big difference in the rate of water evaporation for the 60 and 140 mm fans:

Table 3. Water evaporation rate (ml/hr) for the three fountain setups. Data was determined after 4 days of continuous running for the No fan setup and 2 days for the fan setups.

Based on the results from Table 3, it is clear that the lower temperatures and increased efficiency of the 140 mm fan setup comes at a cost of ~ 35% increase in water evaporation rate compared to that of the 60 mm fan. While “watering” the computer every 2 days instead of 3 days might not be a big deal, when larger amounts of water are used (say 50 liters), it could be the difference between refilling the bucket every 24 days instead of every 17 days.

That is a whole extra week of maintenance-free computing. That alone should be enough reason for choosing the 60 mm fan setup instead of the 140 mm fan, but if you also consider that the 140mm fan is only marginally more effective at cooling than the 60mm, then the 60 mm is a no brainer.

Also of some concern is the greater increase in humidity caused by the 140mm fan compared to that of the 60mm. Originally I wanted to test the fluctuation of humidity as well; however, I quickly realized that it would be very difficult to test it at this time of the year because the windows are open almost every day. I guess I’ll do some testing during the winter when we are forced to keep the windows closed.

The last test I conducted was an overall noise test (Table 4):

Table 4. Noise level recorded using a RadioShack sound level meter set at C in fast response. Noise test was performed by standing the recorder in a tripod located about 8 inches from the computer and in the center of the 11′ X 11′ room.
*Overall room noise after modification of the 2 loudest computers in the room.

This is what I had in mind when I decided to build a water-cooling system. Noise reduction was a must because I have 3 computers in the same room and the Duron @ 1GHz was the greatest offender, with its Global Win heatsink and fan. As we can see, the noise level of the fountain is very low compared to the heatsink + fan solution. Of course, lowering the noise level of the Duron system alone would have not solved the overall noise problem.

So, in order to lower the overall sound in the room even more, I quickly modified one of the other computers (overclocked Athlon @ 750 Mhz) by lowering the voltage of the CPU fan and rearranging the case fan to compensate for the decreased air flow of the lower voltage CPU fan. (It worked quite well, although the details of the modifications are outside the scope of this article.)

After all was said and done, I was able to lower the ambient noise from 62 dB to 50 dB, which was one of the main objectives of the project.
{mospagebreak}

Bruno M. Pancorbo and Paula K. Knabb

Although the data shows the fountain as a good solution for cooling, resolving some little problems was neccessary before it could be a workable solution.

First of all was the water reservoir problem; while refilling the container with 7 liters of water every 3 days is fine for testing purposes, I don’t really think it is a good solution in the long run. For this reason, a larger container was used in the final setup. With the current container

I can run the computer for at least 19 days before refilling it.

And because the computer is idle most of the time and its water evaporation rate is lower (~ 40 ml/hr when the computer is in standby), I can stretch the “watering” to nearly 40 days before refilling, which equates to about 9 times per year in the “Best Case” scenario. I also would like to mention that increasing the size of the water reservoir from 10L to 40L did not nullify the results from Table 1. The only thing that changed was the time to reach temperature equilibrium.

The other problem was how to devise an alarm system that would be triggered when the water level was low, thus decreasing the need for constant monitoring of the system. For this I chose a simple audible door alarm system:

This alarm system has two parts that, when separated, emit a loud sound. By tying a nylon string from one of the alarm system parts to a counter weight that rests on a floater in the reservoir, I am able to trigger the alarm when the water level is lower than the length of the string. The system is crude, but it seems to work well during testing.

The final problem was how to eliminate the growth of unwanted bacteria and algae. Although I initially used plain table salt, I was not happy with its crystallization around the sides of the fountain and in other places where water had splashed. I then toyed with the idea of a weak bleach solution, but I was not very comfortable with that idea either.

Looking in the Forums, I noticed a post about humidifier antibacterial solutions:

Since the waterfall is an evaporative humidifier, it seemed to be the most reasonable solution. Although I haven’t tested it for a long time, I hope it works well. For the chemists out there, the active ingredients of this solution are dimethyl ethylbenzyl ammonium chlorides 2.25% and dimethyl benzyl ammonium chlorides 2.25%.

Finally, even though I used a purchased fountain which cost about $100 (including the pump), I do not believe it would be difficult or expensive to make a similar tower from scratch following the schematics shown above. Those of artistic leaning could even create a more elaborate, aesthetically pleasing tower (some ideas have included using sea shells, glass bottles, rocks and ceramic pots). A trip to the local gardening store should provide inspiration.

If I learned something from my experiments, it was that you really don’t need to keep the water much cooler than RT to achieve good results. In fact, the price for getting the water below RT might be very high in terms of evaporation rate.

Because reservoir volume is a big consideration for most water cooling systems of this type, this factor should not be overlooked. I think a case can be made for not overcooling the water with large fans (perhaps I should test a 30 mm fan?).

Also, it seems that the only way to minimize the difference between water temperature and CPU temperature is either by increasing the water flow rate (just like using a faster fan in a standard cooling rig) or improving the heat block (just like using a better heatsink design.)

Not only is it capable of cooling the Duron @ 1 GHz, but it also lowers the general noise level in the room to more acceptable levels. As a bonus, it works as a humidifier which will come handy during the dry winter days.

Although it does require more maintenance than a heatsink + fan solution, in the end I think the advantages of a water-cooled system outweigh its disadvantages. Oh, and did I mention that it also allows you to run CPU’s at much higher frequencies than they were intended?

Bruno M. Pancorbo and Paula K. Knabb

PS: The following pics show the finished fountain with halogen light accessories: