Check out our water cooling gear!

Water Cooling Project - Part 1

Date: 8/13/00

Go to : Water Cooling Project - Part 2  

Description  

I've been overclocking my PIII500E for the past 6 months and I haven't done any fun projects in a while so I decided to build a water cooled system. Actually, I heard about water cooling many months ago, but never considered it seriously - hey man, do I really want water splashing around inside my computer case? But, it looks like water cooling isn't so crazy judging by all the articles at Overclockers.com; I've perused all the water cooling articles and I owe them for many of the ideas here. Armed with sufficient knowledge to be dangerous, I felt ready to give it a shot and build a water cooled system of my own. Now before I go on, let me make it clear that I don't advise anyone to do this as it could result in serious damage to your computer system, loss of data, or even worse things which are too unspeakable to mention here. As always do your research and know what you're getting yourself into! Safety first kids. Being a little skiddish about water floating around in my computer I decided to build a system that would keep the waterbox, and radiator outside of my case. This is cool since it keeps my case a little less cluttered and also makes for a portable water cooling system. First off, here's a picture of the completed system. I kept the case cover off so you can see the plumbing.

Here's the list of materials that I used to construct the system:

System Diagram

A - WaterPump, B - Copper CPU Water Cooler, C - Radiator, D - Fan. Arrows indicate the direction of water flow.

Here are some pictures of the assembled unit:

Copper Water Cooling Block

I built the water cooling block by snarfing Surleyjoe's design. It's basically a 2" copper cap soldered onto a 2.5" x 2.5" copper plate. I soldered 14, 1.25"h x 3/16"d copper pins onto the bottom of the plate. What a pain that was. I arranged the pins in a staggered pattern on a grid marked off on the plate. I then drilled holes at each pin position so I could set the pins in them when I soldered them. Well, to get the pins to fit in the holes, I had to file the end of each one as I couldn't drill deep enough w/o going through the 3/16" copper plate. Also, since I have only a cheap propane torch (which I use for the occasional job to sweat copper pipes) -I didn't have exactly a surgeon's detailed instrument here. The first pin went in fine and soldered real nice. As I tried the second pin, the 1st pin would also heat up; the solder would melt and fall over (like a domino). Yikes! So, I had this continuous problem of re-melting the solder on the pins. The only way I could figure to solve this was to take some copper electrical wire and run it all around the pins that were in place. This kept them from falling over. It would have better to use smaller diameter pins and that way I could have drilled deeper holes (relative to the diameter of the pin) to set them.

The next trick was to make two holes into the 2" cap for the water in and out pipes. I drilled a couple of 1/4" holes for the copper pipe. I used about 1-1/4" pipe and soldered it in place without too much trouble. I put them off-center from each other, hoping to achieve a swirling effect in the water.

Finally, I positioned the 2" cap on the plate and soldered it on. This was a bit tricky since I had to remove the wire that was holding the pins in place. Fortunately, (I think) all the pins held when I soldered the cap on. Here's the final result. It's a heavy sucker (heavier than any air cooler heatsink I've had).

A couple more things about the water block. Notice that I needed to whittle away some of the copper. This was because of two issues. First, the notch on the back was created so the water block clip would clear the copper plate. I didn't actually plan this, but it turned out to be a good thing since it gives the water block a little more stability with the clip in the notch. The second area I had a problem was with the flip chip height. Since the chip has a lower profile than the PPGA, a flat heatsink will run up against the socket 370 plastic header and there's no way for the copper plate to lay flat onto the chip. This isn't a problem with the PPGA Celeron, but it sure is a problem with the PIII 500E. To rectify the situation, I pulled out my hand drill and sand paper disc and started sanding away until I had taken away enough material to clear the header of the socket 370. This also worked out in the end as the copper butts up flush with the socket and prevents it from rocking side to side.

I did some hunting around for a suitable case and it turns out that file cases work out great for the job. I grabbed this case from Office Max for about $11US. Using my trusty Black and Decker saw, I cut a hole in the back about 1/2" less around than the radiator I am using. The radiator was attached using the mounting clips that were supplied with it. All I needed to do was drill 2 sets of holes on each side of the hole that was cut out and then run the fasteners through.

Note the hole drilled for the power cord.

As I mentioned above, I used a 4"x4"x4" electrical box for the water reservoir. This is the source of the cooling water for the system. These boxes are intended for underground electrical work, but also make a great box for the water reservoir and pump. I drilled two holes for the input and output hose barbs, 1 hole for the power cord to the pump, and 1 hole for screw on cap to fill the reservoir. The filling hole is a great idea I lifted from Surleyjoe.

Mounting Brackets

I made the above pictured mounting brackets from the following materials:

• plastic drive covers from my computer case
• 2 x 3/16 machine screws
• 2 x 3/16 fly nuts
• 3/16 nylon washers
• 2 x compression springs (I think I used #5 springs - if that means anything :)
• Intel PIII FC-PGA retail heatsink clip

Basically, this design is a spring compression design in which I've used epoxy to affix the intel heatsink clip ends to the screws. To do this, I first snipped off the screw heads. Next, I took my intel heatsink clip (which wasn't being used) and using my propane torch, I heated the clip up until it was a bright red and then I carefully bent the clip (with pliers of coarse) so that it was as flat as I could make it. Once, I bended and hammered the clip to be as flat as possible I snipped off the two ends. Then I took this wonderful 5 minute epoxy glue and stuck the clips onto the ends of the screws with about 3/4" overlap onto the screw. I applied a second coating of epoxy with a small piece of fiber glass cloth wrapped around the clip and screw to give it extra reinforcement. Be very gentle with the fiber glass cloth as it is made of very loose strands and will fall apart easily. Here's a close up of the entire clip:

One point of caution is that I have created a battery here. This is because we're mixing copper and aluminum with our friend water - which if you remember your high school science - is an electrical conductor. This effect is called electrolysis and can eat away your metals. In order to avoid this, I've added a product known as Red Line Water Wetter, which you can pick up at your local auto store. This stuff is supposed to slow down corrosion, and it supposedly helps with cooling to boot. Don't know if it works or not. You may notice the slightly pink color of the water in my hoses (that doesn't sound right), well that's Water Wetter.  

Fill'er Up!  

I bought a gallon of distilled water from the local grocery store and added the Water Wetter to it. Then using a funnel, I filled up the reservoir through the "dead lock". When the reservoir was about 1/2 full, I turned the pump on and kept filling until all of the air was out of the system. One point here, the copper water block may have some air trapped in it, so before you clamp it down to the CPU, give it a couple of turns to free the air bubbles. I tried to fill the system with as much water as it could take without overflowing; I figure the more water in the system the more difficult it will be to heat up.  

The Results?  

The system is installed and running. No leaks (yet). I'm currently running at CPU temp of 28C for over 3 hours under normal operating conditions. I plan to do some benchmarks in the next few days and I'll post the results here.  I've been leaving my machine booted and running continuously - 24x7. The numbers in the table were collected over time by monitoring HWDoctor for CPU and System Temps, several samples were taken and averaged.

Testing Machine:

• ABIT BE6 II Motherboard  
• Pentium III 500E (FC-PGA)  
• MSI-6905 Master Slocket  
• 256MB PC133MHz SDRAM  
• Leadtek WinFast Geforce 32MB DDR Rev B

I decided to add quick connects to lessen the hassle of unscrewing clamps and pulling hoses off the barbs every time I need to do some work on the computer. This makes it easy to disconnect and connect. Here's a picture showing how they're put together. Plumbers (Teflon) tape was required to keep the threaded pieces from leaking. These are heavy and can cause the tubing to be pinched and restrict flow. To avoid this problem, I placed them far enough along the hose so that it wouldn't be putting weight on the hose and cutting off flow (they rest on the floor).

Alternate Water Block

I've constructed a second waterblock as an alternative to the copper waterblock I used above. This waterblock is made from plastic, an aluminum heatsink (Global CPM25603-32), brass barbs, and the liberal application of JB Weld adhesive. Here are some pics of the heatsink:

I picked up a sheet of clear Plexiglas from (where else?) Home Depot. I ran rows of masking tape over the surface of the Plexiglas to give me a surface to draft cutting lines on that would stay in place as I cut the plastic. I sized the panels such that the sides and bottom panels are glued directly to the heatsink itself using JB Weld. I tried using Plexiglas for the hose barbs but it was too rigid and it cracked when trying to thread the barbs through. I hunted around the house and found a plastic lid from a bucket of dry wall compound that looked good for the job. The lid is the blue plastic in the above pics. This plastic is more flexible and tupper-ware like, which made a nice tight fit with the hose barb threads. 

The first time through this, I didn't apply very much JBW and the box leaked like the Titanic when I hooked it up to the pump. Another mistake I made was not to roughen up the plastic surfaces. For both of those reasons, the box wasn't water tight and in fact, the glue didn't take a strong hold. I ended up prying off the panels (which came apart clean and easy) and started over. This time, I scratched up the edges that would be glued to one another and I was extremely generous in the application of glue. As you can see in the pictures above, it's not exactly the neatest job ever, but it seems to hold it's water now.

The nice thing about this block (provided it doesn't leak!) is that I've been able to use a good ALUMINUM heatsink.  This is good, since I can avoid the problem of corrosion caused by electrolysis. Another benefit is that the heatsink is a commercial heatsink that of course has a lot of surface area for cooling.

If I'm reasonably confident that this sucker won't leak, I'll run some benchmarks and throw them up on this page for comparison to the copper waterblock.

UPDATE: Couldn't get this block to seal tight enough so I gave up. But I had more luck in part 2.

Continue to Part 2: Water Cooling Project - Part 2

Article by N.Curtiss, last updated: September 10, 2000.