AquaStealth – How To Mix Water and Silicon


About three weeks ago I did the review on the AquaStealth CPU water cooler (below). I have been running it since I first tested it and I thought it would be interesting to revisit it.

I am using it on my main system – I call this my “work machine” and don’t fool around with it too much. So for me reliability, no fuss and quiet (I use this system on average 4-6 hours/day) is really important. When I test heatsinks and other stuff, I use this machine but never in a way which would jeopardize system integrity. For the “far out” stuff, I swap out hard drives and use a small drive that contains nothing important, so crashing the registry is not an issue.

What I wanted to monitor during this time were key temperatures – CPU, air intake, temp at the radiator and exhaust temp. I figured that these would give a good idea of how this unit behaved. What I found was the following:

    o The temp difference between intake air and air at the radiator averaged less than 1 C;
    o The CPU temp (thermal diode, PII300@450) averaged 1 C below ambient at rest, never more than 6 C over ambient under stress;
    o Exhaust temp averaged 3 C over intake air temp, exhaust measured at the power supply fan exit.

This is impressive and consistent performance, and all very quiet compared to running the same system with an Alpha or any high cfm two fan heatsink. The water pump is inaudible when the system is on – the 120 mm fan is rated at about 30 dBA and the power supply fan is also very quiet. I have two other fans, one Lasagna cooler for the BX chip, and one 50 mm aimed at the power transistors behind the CPU.

I have surrounded the water pump reservoir with bubble foam to cut the noise and it works great. Even with the system off and the water pump running, it is almost inaudible at 3 feet from the case. I have my case inside a cubby hole in my desk, and with this arrangement I doubt the noise level gets over 30 dBA. I am mightily pleased.

Water cooling at least an order of magnitude more efficient than air and as I found with peltiers numerous times, it is a superior way to get super cooling without horrendous noise and massive heatsinks. Water cooling is going “mainstream” and you will see additional products from “serious” companies looking for volume sales.

UPDATE 11/11/99: AquaStealth With Tom Leufkens’s Peltier Kit

SUMMARY: The AquaStealth with Tom Leufken’s Peltier Kit can make your water cooled peltier dreams a reality for about $130.

There is a lot of interest in using water cooled peltiers and the AquaStealth seems like a natural base to build one. I used Tom’s Socket 7 Peltier Kit to see how the AquaStealth handles the extra heat. I was concerned that the peltier would add a lot of waste heat into the case and make it an intolerable setup. I was pleasantly surprised to find very little heat build-up.

I used the same setup (C366 with slotket) as used below. Tom’s Peltier Kit comes with the peltier, aluminum cold plate and foam gasket. After assembling this sandwich, I wrapped it up with plastic wrap until it looked like a ham sandwich to go (“AquaWrap”).


Because this is a socket 7, I figured (at least for this test) that wrapping it up as tightly as I could would prevent air from reaching the back of the CPU, thus avoiding any condensation problems. Longer term, some creative work with epoxy and fiberglass is probably the best bet.

The table below details the results for the C366 @ 550 and 604, 2.0 volts, with Prime 95:

ConditionCPUIntakeRadiatorCu Block
C366 @ 5506 C19.8 C22.8 C21.3 C
C366 @ 6049 C20.0 C23.5 C22.0 C

CPU is temp using the chip’s thermal diode, Intake is temp at the 120 mm fan’s opening, Radiator is temp at the center rear of the radiator, Cu Block is temp at the center rear of the water jacket; all temps were with Prime 95 running.

This is a 20 C temperature drop from what I measured with water cooling alone. I tried to bump up to 616 MHz but it was a no-go @ 2.0 volts. I’ll do some more experimenting but I have set a goal of seeing what’s possible at spec voltage, although a voltage bump would probably do the trick.


This is about the easiest way to build a water cooled peltier rig I’ve seen. But these are kits – controlling condensation and taming the peltier does require careful attention to the details, but the bragging rights are yours to take.

First Review

SUMMARY: The easiest way to dive into CPU water cooling – for about $100, all the components you need. It works, it’s very effective and could be a great foundation to build a water cooled peltier rig.


Pictured here is the fully assembled AquaStealth (including water) I mounted in my case with the C366 on a slotket (not included).

When I opened the box containing the AquaStealth, my first reaction was “You must be kidding!” It is, to say the least, under-whelming (I am getting a digital camera today – pix will be added soon). I first tried CPU water cooling with Chen’s copper cooler (HERE). As I mounted this in my system, I had to search out and buy a pump, clamps and extra hoses. I then had to set up a cooler chest for the water reservoir and proceeded to dump ice packs into the cooler chest to get sub-zero peltier temps. No question it worked great – no question it was unwieldy and not a permanent cooling solution, but still great fun.

I have always thought that some kind of closed-loop water cooler had the most potential for a viable cooling solution; the AquaStealth is the first “out of the box” product that closes the gap between wish and reality. I am impressed! I am writing this review on my main system with the AquaStealth cooling my C366 @ 550 to 18 C, two degrees under room temp. It is a super quiet, incredibly effective CPU cooler that outshines any air-cooled alternative.


Open the hood of your car and look at the cooling system; what you see is an engine, a radiator and a fan. Under-whelming, but it works very well for your car. Apply the same setup to cooling your CPU and you have the AquaStealth. The kit comes with a radiator, water reservoir, 120 volt 40 gph water pump, hose, clamps, 12 volt 70 cfm 120 mm fan and a copper block water jacket (which could use some lapping). What’s lacking are clamps to mount the water jacket onto the CPU. For this test, I am using rubber bands and a slotket with my C366 on a SOYO 6BA+IV.

One great thing about water cooling compared to air cooling is its efficiency. Water is a lot denser than air and consequently can absorb a lot more heat than air. The reason overclockers are so into high cfm fans is the necessity to move a lot of air to absorb heat. Very simple when you think about it – air can’t absorb a lot of heat, so you’re forced to use a lot of air. Problem is noise and diminishing returns – for a given heatsink, you could increase air flow by ten times and get little or no improvement in cooling. When you hit the wall, you have to consider a more effective heat absorbing medium, like water. Cheap, very efficient but insidious – if water can find a way to get loose, it will. Mixing water with silicon is not good, and this one of the key considerations with water cooling – leaks.


There are 5 possible points from which water might leak – the hose connections and the water reservoir. I put clamps on all connections so the possibility for leakage is nil. The water reservoir looks like a Tupperware container (2 cup size), but seems capable of holding a tight seal. The only problem I can see is if you tip it over. Easy enough to fix – mount a velcro pad on the bottom and stick it to the floor of the case. However, even if you didn’t do this, you would have to do some serious case tipping to knock it over – the hoses keep it in place. Considering the temps involved, there is very little stress on the system aside from water pressure.


Now it does not come with a lot of hardware, so I mounted the reservoir to the fan using tape – temporary of course, OK for testing. Long term you have to use long bolts to mount the radiator to the fan. My case already has a 120 mm fan for air intake, so I did not have to cut any holes to mount the fan. You must cut an intake hole in your case for the fan, otherwise it will not be effective and will be very noisy. The other alternative is to mount the whole thing outside the case. To do this you would have to build some kind of enclosure and run wires and hoses outside the case, but I don’t think it would be a very difficult task.


The following table details how the AquaStealth performed:

ConditionCPUIntakeRadiatorCu Block
C366 @ 55028 C20.3 C21.6 C22.1 C
C366 @ 60429 C20.3 C21.8 C22.9 C

CPU is temp using the chip’s thermal diode, Intake is temp at the 120 mm fan’s opening, Radiator is temp at the center rear of the radiator, Cu Block is temp at the center rear of the water jacket; all temps were with Prime 95 running.

I calculated system cooling efficiency from the recorded data and find a C/W of 0.38, very low* (Methodology go HERE). The best system C/Ws for air cooled solutions measure about 0.6, so the water cooler is very effective indeed. It enabled me to get my C366 to 604 which I was not able to do with air cooled Alphas. In normal use, I have not seen CPU temp go over 23 C. With Waterfall running, CPU temp at rest is 18 C with ambient at 21 C. Particularly note that the difference between the Intake and Radiator temps is less than 1 C. I was concerned about the radiator dumping a lot of waste heat into the case, but with the temp differential so low, this is not a concern. It will be different with a peltier and that is the next installment of this review.


Well, I see a lot of cooling hardware come across the stoop, but the AquaStealth is the first closed-loop CPU water cooler kit that I can recommend. It is a kit and it is not a polished product, so if that’s what you’re expecting you will be disappointed. Make no mistake about what this is – it is put together by “one of our own” (Brian Eisenbrandt developed the AquaStealth). It is however worth $100 if you want to try out water cooling and looks like a good building block to experiment further with water cooled peltiers – more on this later.

*The AquaStealth calculation for C/W is not comparable to what I used and, IMHO, overstates cooling efficiency considerably.

Be the first to comment

Leave a Reply