Passive radiator cooling – Peter Loft
I suppose it all begins with the fact that I don’t like noise. I may rattle around in the wood shop all day long, pounding nails and sending hunks of wood through a screaming table saw, but when it comes time to sit down at the computer and write an article or a few e-mails, it seems that even the smallest sounds annoy the whatever out of me.
My new Athlon 2000XP+ system has been just fine in all departments excepting that of noise. It runs on an Asus motherboard in a case with a silent power supply, an Arctic Cooling “Copper Silent” heat sink on top of the CPU and a final fan cooling the case. It was, admittedly, quieter (as well as MUCH faster) than my old Celeron 433 MHz system, whose aged fan sometimes whined and screamed the displeasure of it’s worn out bearings.
And yet, after reading through all the marvelous articles on the internet about the benefits of water cooling and how quiet they could be, my thoughts and dreams began to be filled with the NON-sounds of a fanless computer with water cooling.
Still, one thing has always puzzled me about most water cooling systems: The designers come up with clever cooling blocks which no longer need a fan sitting on top of them. They then pump the water to a radiator… only then to once again use a fan (if not two or three) to dissipate the heat from that radiator.
Not only that, but most of the systems stick the radiator in the case, so that the heat still has to be exhausted out of the case as a last step with – you guessed it – another fan or two. Now if you are after overclocker’s heaven, I can understand it; but for a simple soul like myself looking for peace and quiet in his office space, it had to be different.
While pondering this question, I stepped into the kitchen to grab a cold, crisp apple from the fridge. And there the inspiration hit me. It was right in front of me! The everyday refrigerator does a marvelous job of dissipating heat from it’s coolant compression system… and without a fan.
It does it passively, through the extended coolant coil system mounted on the back (out of sight). The air passing up and around that rear-mounted coolant coil does all the work. My decision was to try to do the same thing for my computer, to see if a passive system (well half-passive, really, since I need a pump) could dissipate enough heat to run my computer successfully.
However, there were a few problems with this idea – the main one being that, like most of you out there, my ideas are a lot larger than my pocket is deep. I had to make a system on the cheap and that meant that I couldn’t afford to buy any of the fancy, off-the-shelf items that are already tailor made for computers.
Now, I’ve done construction, model-building and hobby work all my life, so there wasn’t to be any trouble on that score, but I had to invest a lot of quality scrounging time before I even started the project. This meant being resourceful and figuring out where to beg, borrow or whatever the bits and pieces to make the system come to form.
I also put in an amount of time doing some mathematics to estimate the surface area that the air coolant unit had and how many times more this is than the tiny, tiny surface area of the Athlon CPU. I ended up deciding that a radiator having approximately 1200 times the surface area of the CPU die would be a good place to start. With that as my main guideline I decided to charge in and begin.
My reservoir began life as a cut-off section from a one-meter length of 8 cm square copper downspout. With a bit of appropriate cutting, folding, bending and soldering, it took form. The finished reservoir measures 13.5 cm long by 8 cm high by 8 cm wide, exclusive of the tubes and dome.
Lacking a ready source of thin sheet copper at the time, I cut out and hammered flat other pieces from that same length of copper downspout to serve as stock for other bits of the project such as the reservoir cover, the back panel water and the pass-through at the back of the computer.
The thickness of the copper from the downspout is .7 mm, and when I needed a thicker piece or reinforcement along the edge of something, I used it doubled or folded over.
On the top right of the reservoir, there are two tubes (10mm diameter). The straight tube is the connection to my length of plastic filler pipe. The tube with the right angle corner serves as the flow return from the radiator.
The lid has folded over edges that were first held down with cut-off copper roofing nails used as rivets (heated to a red glow, cooled off, cut to length and peened over in place). After that, the edges were soldered and drilled. This is probably over-kill, but I like a clean job. The little soldered-on notch as seen on the lower right hand side is the power cord pass-through.
The pump for the system (as pictured) is a Seltz 480 lp model that I picked up at a flea market for almost nothing (it might have been 50 cents, but I don’t rightly remember). It sits within the reservoir with a couple of glued on rubber tabs underneath, connected via the short length of 12 mm ID tubing (the green bit on top) to the inside of the copper feed dome in the reservoir lid. The flow from the reservoir is fed directly from the tube atop the dome up to the waterblock.
The lid is held in place by eight brass screws and the seal was effected Loctite Blue Silicone RTV Form-A-Gasket material. This forms a water tight seal that is good up to 250° Celsius and is impervious to water, alcohol, oils and various other nasties. This material made for a perfect water-tight seal the first time through.
It was while finishing up the water reservoir that the name for the system came to me. The clunky, old-time look of the reservoir reminds me of nothing so much as a piece of a steam engine and, for no particular reason, the steam engine that powered the boat in which Humphry Bogart and Katherine Hepburn travel down river in the classic film, “The African Queen”. And so the name stuck.
I was lucky where the material for the water block was concerned.
A friend of mine who works in the design department of a local company was able to get a few short cut-offs of copper bars from the workmen on the factory floor. The body of the waterblock is made from bar stock; 10 mm thick by 40 mm wide cut to a length of 60 mm.
I have studied as many of the articles on the internet as could be found regarding water block design and construction for computers. I chose to use a double water channel design to try to achieve a maximal flow rate through the waterblock while at the same time using as much of the waterblock’s area.
The design is, of course, something of a compromise.
Since I have no access to sophisticated metal milling machines and such, I opted for a design that could be created by drilling rows of holes on my drill press. I drilled the rows of holes using a 6mm drill, but used larger drills (even up to 10.5 mm) at the corners where the water changes direction to try to reduce the flow resistance.
The channels are drilled to a depth of about 6.5 mm which, by my calculations, allows the twin channel design to have about the same cross-sectional area as the minimum diameter of the copper pipes in the system, this again to minimize flow resistance. The drilled holes were connected first using my Dremel rotary tool with a carbide cutter and then by using a series of fine machinist’s files to smooth out the channels a bit more.
The block was finished off by soldering on a 3 mm plate into which are fitted the 10 mm inlet and outlet tubes. The design is such that the coolant flow from the pump is admitted at the area marked “A” in the middle of the waterblock and travels around via the twin channels and then exits in the corner of the block at “B”.
The waterblock was assembled, as were the other parts in the system, using an ordinary low-temperature lead solder. It seemed unnecessary to use a harder, high-temperature solder for this application since even a temperature that would be hot enough to ruin the processor and stop the whole system would still not be hot enough to affect the strength the solder joins.
When the soldering was finished, the waterblock was tested for leaks and then lapped on the bottom using progressively finer grades of waterproof sandpaper on a sheet of glass (the finest being something like 1200 grit) to achieve a serviceable, reasonably-flat surface to contact the top of the processor.
The following image presents the clip system used to hold the waterblock securely on top of the processor in the socket:
As I was not able to find springs that would serve, either flat or in reasonable small coil diameters, I opted for a screw-on approach. I used double nuts on a threaded shaft clip and a set of washers with a somewhat soft bit of rubber between them acting as a buffer to allow for limited expansion of the waterblock as it heats up in use.
As the clip system must take a bit more tension than any of the other parts of the cooling system, I opted for brass instead of copper here, using a jeweler’s saw to cut the holes, again finishing off with very fine files and then soldering the parts together. It is difficult to make out in the picture, but the side clips also have a tiny little rivet (made from a short iron nail) to assist in holding the threaded rod to the socket base clip, as I did not want to have to completely trust to the soft soldered joint for this purpose.
Here is the assembled clip system on the waterblock prior to installation:
And here is the block, firmly attached in position with the clips waiting for the installation of the hoses:
The cutout in the brass strap was necessary so that the strap would fit across the geometric middle of the waterblock and, in so doing, bring the tension down properly on the top of the CPU.
I used a no-name heat transfer paste and then carefully screwed down the nuts to apply what I felt was “just” enough pressure on the block (to the point where it could no longer be shifted on the die by pushing on it). Then while holding those nuts with a tiny wrench, I tightened down the lock nut on top of it with a second wrench.
As unscientific as this might be, nothing went “crunch” or “snap”. A side benefit of the cutout in the strap is that it holds the waterblock in place quite nicely so that it doesn’t fall during the tightening up procedure.
Here is a close up of the reinforced plastic tubing I selected to use with the system:
It is actually designed for commercial water installations in restaurants and such. Sharp eyes will note an absence of clamps where the hose meets the copper tubes in this system. In truth, the clamps are completely unnecessary.
I made sure to leave enough length of copper tube so that the hose would have sufficient contact (no short stubs here). The hose has such a good fit on the copper tube that when found I had assembled one of the connections incorrectly, it took a pair of pliers on the hose as well as another pair of pliers on the copper for me to pull them apart. I haven’t seen so much as a single drop of a leak from the system since day one.
Here is the mounted waterblock now properly plumbed and ready for use:
This plastic hosing was certainly rigid enough so that I had no fears of it kinking. On the other hand, I thought that the bends of the hose from one mounting point to another might add undesirable side loading or strains on the waterblock mounting.
To get around this problem as well as to route the water more cleanly, I ended up using the tubing within the computer more as a flexible coupling for lengths of copper pipe that I custom bent to run from reservoir to waterblock, waterblock to back panel and so forth. Here are pictured some of the various bits of copper tubing, shaped and ready for installation. The little stubs on the right side are simply sealed bits of tubes to be used as stoppers for the filler hoses.
Here is a photo of the internal runs of tubing within the PC:
The filler tube is normally left standing inside the case along the back edge, but was bent down out of the way so that the details of the plumbing might more easily be seen. The pump is plugged into a power strip that provides current to the computer, screen and printer. I have to turn on that strip to turn on the computer so that the cooling system is always the first thing to come to life.
Here is a close-up of my back-panel coolant pass-through strip:
I double-layered my copper, riveted it solidly (three tiny iron rivets), soldered it tight and then cut, bent and drilled it to form. The 10 mm pass-through tubes are soldered in place with short lengths of 12 mm tube around them for stability and strength.
Here is the radiator for the system:
I think that this is currently the weakest link in the system. I decided on this method of assembly as the straight copper tubing and corner pieces seemed a cheaper answer than buying an expensive tubing bender for use with flexible copper tube.
These are 15 mm diameter copper tubes and the two and one-half meters of length provides a fair amount of surface area to dissipate the heat from the coolant. I had intended to solder on copper wires to connect the top surfaces of all the pipes as a crude form of finning, but could not generate enough heat with my tiny hand-held propane burner to actually bring that part of the design to form.
The back view of my computer shows the hose connections between the back panel and the radiator:
The tee-connection at the top left is actually another filler point for the system (air vent), but I decided to double-up on its utility by employing it as a handy mounting point for a photographer’s thermometer with which to measure coolant temperature.
A close-up of the thermometer tells it all:
Supposedly this photo thermometer is accurate to plus or minus one-half of a degree. The coolant temperature maxes out at 100° Fahrenheit (37-38° Celsius). It never seems to get either hotter or colder.
Finally, we see a side view of the computer with the radiator in place:
I had intended to make a fancier mounting arrangement than just that bit of iron wire, but since I am thinking of finding or creating a better radiator for this system ,I haven’t done anything to clean it up or make it more secure.
As with most of my projects, I finished it up very late one evening and decided to test it on the spot.
The system was filled with a mixture or distilled water, a few dashes of methyl alcohol and just a touch of methylene blue microscope slide stain for color and anti-bacterial properties. I got my “distilled water” from my dehumidifier, since water that is condensed from out of the air won’t have any minerals in it.
It was only after all the trapped air bubbles were finally out of the system and the pump had been running on the bench for about thirty minutes to check for leaks that I dared start up the computer, itself. Everything went just fine.
There wasn’t much noise anymore, but would it work? Would this cooling system hold the CPU to a reasonable temperature, or would it just end up as an expensive bit of fried, high-tech circuitry?
Actually, it worked quite well.
I thought that the system would just get hotter and hotter until it ran away with itself, but it all leveled out after about an hour and a half at around 47° Celsius. The external radiator feels pretty warm, but then that is its function: To get warm and dissipate the heat.
My only means of reading out the temperature of the CPU and motherboard are by means of the Asus Utility “Probe” that interprets the signal from the motherboard. I have no idea as to whether the temperature sensor built into the motherboard is actually “calibrated” and accurate, but in any case I can give the “relative” performance temperature readouts between the computer system in its air-cooled state and the system in it’s water-cooled state.
45°C / 113°F
52°C / 125°F
39°C / 102°F
47°C / 116°F
50°C / 122°F
33°C / 91°F
What can be seen here is that the experimental, half-passive water cooling system in its current condition is essentially as effective as the air cooling system I had been using on my system. It can also be seen that there is a side benefit, in that the motherboard is running much cooler than before – this undoubtedly since it no longer has a constant stream of hot air from the CPU being blasted down on it by a heat-sink fan.
I am planning to test another radiator on the system in the near future, but have to first try to find some information that might indicate how much cooling improvement I might gain with a certain percentage of extra surface area on the external radiator.
In the meantime, I’ll just have to content myself with a system that is so quiet that I can hear the birds chirping in my garden again, where the hard drive is louder than the cooling system.
And am I proud? You bet! I called my wife into the room the next day and said, “Here it is, Dear, my new silent, water-cooled computer system. I call it the African Queen.”
She took a peek inside at the maze of shiny copper and plastic hose and quipped “Seems to me like you should have named it the Borg Queen!”
Oh well, you can’t win ’em all.