The 'Convection Plant' Prototype `-Updated 2007.07.13`

-Who said it isn't possible?

-A working prototype fanless, pumpless, passively radiated computer running an AMD Athlon xp2000 CPU at standard core voltage.

Eat your heart out, Heath Robinson (pas-cool-main.jpg)

More pictures:

Overview of the Cooling Rig

Close-up of the motherboard / Waterblock

The Reservoir

An 'Action' Shot -like watching paint dry :)

and finally, it's Coffee-break Time...

Background Thoughts:

If you wonder what the point of this system is, it's for silent cooling; that's all. People who overclock computers won't be impressed by what they read here. People who want their computer to be quiet *might*. I say *might*' because for most people, careful choice of computer parts avoids the problem and there is always the option of buying off-the-shelf 'quiet bits':- low noise cases, special fans, passive heatsinks for graphics cards etc. Total silence is hard to achieve though and most people find some compromise between noise, cost and effort involved which suits them. It is obvious from the number of silent computing websites, fora and companies selling 'low noise' computer parts that lots of people want to reduce the noise of their computers, and as with anything that lots of people take an interest in, some take it to extremes. For these 'computer silence extremists', this web page should be of interest. These people are prepared to sacrifice more money, time and effort in order to push harder for low noise and will stop at nothing (lol).

I have been interested by the possibility of a water cooled computer which uses just natural convection heat transfer for a while now. I searched the web for a whole day to see if anyone had done the same thing. -sad eh? :) All I found was a few systems that had passive radiators and another, more interesting setup using an AMD K6 processor (see links below). I also, rather ominously, found a thread in a forum where someone asked if pumpless water cooling was possible; the poster was snubbed by a series of replies pouring scorn on the idea and dismissing it as impossible.

The K6 setup was really close to what I wanted, but I couldn't be sure if fluid convection could keep up with a more powerful processor or whether the open mixing bowl of water balanced on top of this guy's (open-sided) tower computer (-I kid you not) was providing most of the cooling by evaporative loss.

If I hadn't lost the link to this website, I suppose I could have emailed the builder to ask these questions. Instead I decided to build a similar rig myself.

I thought about the possible problems facing a pumpless, convection-driven cooling system and realised that if it was going to work then I would have to bear in mind the old maxim: 'Compromise is the Enemy of Achievement'.

Most of the above rules seem like common sense but with a passive convection system, there is very little incentive for the water to flow and the problem of friction becomes serious. The 'radiator above the CPU' rule came from an Israeli website which suggested this was good practice. (The website was referring to a pumpless cooling system for buildings -not as good as air conditioning, but dirt-cheap to build, run and maintain and *certainly* better than no cooling at all).

I made some design rules:-

Make the water-flow as straight as possible.
Where a change of flow direction is needed, give the water plenty of room to make the turn.
Make the pipes large diameter to reduce friction.
Eliminate all restrictions smaller than the pipe diameter.
Radius the entrance of holes to minimise restrictive turbulence.
Place the radiator above the CPU.
Take the extra trouble to make special adapters for the pipes to avoid possible leakage problems.

Most of these rules seem like common sense but with a pumpless system there is very little incentive for the water to flow and the problem of friction becomes serious.

Construction:

I went to a local model shop to see what they had to offer in the way of piping. Dispite the distraction of all the cars, boats 'planes, tanks and electric helicopters etc. I found some silicone rubber hose with an internal diameter of 3/4" (19mm) -BIG STUFF! They had two short pieces so I bought them both.

Back at work I found some off-cuts of 1" dia. aluminium rod. I stayed late one night and turned up four barbed tubes to accept the silicon tubing. The lathe I was using to make these had no coolant and, being late, the works stores was shut. In desperation, I used diesel (bad idea / may I be forgiven). Although it's not unusual to use paraffin as a lubricant for drilling holes in aluminium, this was different, it smoked a lot, coated all the lathe controls, making them slippery and generally stank the place out. Having cleaned the mess up, I took my components home.

I got to work on the computer and started drawing up the waterblock using AutoCAD LT. I like to draw things I make in CAD first, but everyone has their own way of working.
Screen-dump: of the CAD drawing

All the main parts of the waterblock were made of 3mm thick aluminium. This could be seen as overkill, but I wanted the heat from the base of the waterblock to flow all round so the water would get heated from all sides.
Photo: waterblock components

After a bit more 'filing and smiling', the waterblock components were ready for T.I.G. welding. I went to see our best welder and pleaded... 'Dave', the welder was inspired and took pity on my cause. During his lunch-break, he tacked the bits together in the corners. This gave me the option of trying the waterblock on the computer before final welding. (If it hadn't fitted, I could have easily broken the light tack welds off to make an adjustment.)
Photo: weld tacks
Photo: tacked waterblock

For those not familiar with T.I.G. welding, it's brilliant, producing clean, strong and dare I say, beautiful welds. A Tungsten tip, shrouded by an Inert Gas heats the metal with an electric arc, while a separate filler rod is melted into the molten pool. The problems with TIG welding are that the equipment is expensive and it takes a fair amount of skill (OK, so that's an understatement) to do well. I've tried TIG welding on some scrap metal and the result wasn't pretty.

With my new-found respect for TIG welders, I tried to take a photo of Dave for the web page, but being a modest chap, he kept turning away from the camera and in between laughs, he used lots of bad words to discourage me. In the end I had to give up. I will be buying Dave coffees from now until Christmas for this favour. Cheers, mate!

In the link below, you can see inside the waterblock.
Photo: waterblock inside

Testing:

The Prime95 program was used to stress the CPU and after 4 hours of running, the temperatures stabilized fully. The specs. below tell the rest of the story pretty well.

Cooling System Specs:

Waterblock: Custom made from aluminium with 3/4" (19mm) internal diameter pipes
Tubing: 3/4" (19mm) silicone rubber hose (from a model shop)
Radiator: Air to air intercooler taken from an Opel Vectra turbo diesel car (matrix dimensions of 5"x3"x9" or 130mm x 75mm x 235mm)
Pump: What do you mean? -I didn't use one. [grin]
Fan: No, no fans here either.

Computer Specs:

ECS K7S5A motherboard
128 MB PC2100 RAM
AMD Athlon xp2000 Thoroughbred core CPU (run at 1700MHz, 1.5v)
Asus V8200 GeForce3 Ti200 graphics card
Quantum Maverick 2GB HDD
AOpen 250W PSU
Windows 98 OSR2
Prime95 'Torture Test' to stress the CPU.

Temperature Measurement:

Remote Probe: Separate battery powered LCD readout with thermistor sensor.
CPU: Motherboard thermistor (mounted below the CPU)

Results:

CPU temp. = 55deg. Celsius
Waterblock temp. = 43.7deg. C (measured at the base)
Radiator inlet temp. = 33.3deg. C. (measured at the top)
Radiator Outlet temp. = 31.8deg. C. (measured at the base)
Ambient temp. = 25deg. C. (monitored by LCD thermometer)

I watched small specks of dirt through the translucent pipes which travelled round at about 0.3m/s. I did a calculation and reckon I am getting around 300 litres per hour of water flow. This is just a crude measurement but it gives a rough idea of what the flow is. In case you are wondering how I calculated this figure, I include the calculations below, -just remember that I am no mathematician and probably even worse at explaining math. If you want to email me pointing out mistakes, be gentle with me :)

Flow Calculations:

Constants: Pi = 3.1416 (approx.) 1 Hour = 3600 seconds 1 litre = 1000000mm³ Coolant Pipe Internal Diameter = 19mm Observed flow speed = 300mm per 1 second interval

Derived Terms: Radius = Diameter / 2 Area = Pi x (Radius²) Volume = Area x Length Pipe Internal Area = 283.5mm² (Derivation: Pi x 9.5 x 9.5) Volume Flow = 85050mm³ per second (Derivation: 283.5 x 300 = 85050) Volume flow = 0.08505 litres per second (Derivation: 85050 / 1000000 = 0.08505) Volume flow = 306 litres per hour (Derivation: 0.08505 x 3600 = 306)

So calculated coolant flow is 306 litres per hour. Let's say (for the sake of argument) that my measurement of the speed is accurate to plus or minus 20 percent. This would give flow rates of 245 l/h (min.) or 367 l/h (max.)

Conclusion:

Before I go on, I must point out that I did not have good contact between the CPU and waterblock during the test. The cause was insufficient pressure applied by the springs that held the (heavy) waterblock in place. There wasn't much I could do about this at the time and I have chosen to report my tests despite this flaw.

Looking at the figures, you can see the big difference in temperature between the CPU and waterblock base. This shows that poor contact is occurring and suggests that the CPU temp would be lower if there was better contact. It also looks as if the waterblock might not be very good at transferring the heat to the water as there is a big temperature difference between the waterblock base and the temperature of the hot water outlet. Perhaps it would work better if there were fins on the inside, as suggested by a friend of mine. The radiator temp. is not very hot, suggesting that it is not the weak point in the system at this stage.

This has been an interesting exercise, and I think this system is worth integrating into my main computer. I would happily run my main computer at this temperature and for people who like silent computers, the elimination of a processor cooling fan is a big step forward on the road to a silent computer. I need to do some more research into waterblocks and the means of attaching them properly.

Future:

I will build a test rig using a high wattage resistor to apply about 75 Watts of heat to the waterblock. In this way I can test different components without involving a computer.

I need a new waterblock for my main computer because the CPU socket is mounted a different way round to my old K7S5A mobo. With the help of the test rig, I should be able to develop something even better than the prototype.

I already have a few other 'quiet bits' fitted to my main computer so I am fully prepared for the day when I integrate a totally passive water cooling system into it. I have a Silentmaxx Fanless 350PCS (350Watt) ATX PSU, -it's very quiet and an expensive luxury, still, they have come down in price since they were first sold. I also have my Seagate Barracuda IV hard disk 'imprisoned' in a diecast metal box, with the wire holes sealed off with silicone sealant. Finally, I have a Zalman passive heatsink for my graphics card.

Links:

http://www.folgorante.net/sirpal/sirpal.html -A Totally Passive-cooled AMD K6 CPU site ... with photos and thermal calcs. (and my project inspiration) -Thanks, Marco!

http://forums.silentpcreview.com/ -Silent PC Review Forum ... (web based forum discussing nearly all aspects of PC silencing)

http://www.plees.f2s.com/ec/test-rig/test-rig.htm -My Research into Passive Watercooling ... Experimentation into the best way to Watercool a PC without a pump or fan.