Vapor Phase Change Cooling

This project has been many months in the making and I finally had enough time to put this thing together. For those people that aren’t HVAC engineers, Phase Cooling involves using a refrigeration compressor to compress a gas till it changes phases to a liquid. This liquefied gas is then pumped into an evaporator which is place on your processor/GPU. While inside the evaporator, the liquefied gas changes phases back to a gas, and while doing this it sucks up heat and gets pulled back into the compressor, completing the cycle. Now that you have been educated in the basic principles of phase cooling, it is time to get into the meat of this article.

Update 10/11/2007

I just realized this guide is terribly out of date and terrible organized so I’ll be working it over the next month and photographing my current builds to provide some stock photos and videos of the assembly process. Just rolled out the first unit of my new line, feel free to check it out here.

Phase Cooling 101

Now if you are scratching your head and feel very uninformed about this topic, don’t feel bad, it took me almost a year to learn everything. To start with, I have a handy diagram that will help you understand the physical layout of one of this Single Stage (SS) units.

Starting at the Compressor, the gas will be compressed to the operating pressure between 150 and 250 PSI at a whopping 70 degrees Celsius. From the compressor, this hot gas is fed into a condenser, which is essentially a radiator for gases. Inside this condenser, the temperature drops to around ambient, for me being 27 degrees Celsius. While inside the condenser, some of the gas may start to liquefy, but the majority of the gas will still be in a gaseous state. Upon exiting the condenser, this mixture is forced into something called capillary tubing, cap tube for short. This cap tube has a diameter between .6 and .9 millimeters, and a measured length, which will be discussed later. Inside this cap tube, the mix of gas and liquids gets compressed till it is mostly liquefied gas with a few bubbles in it. This cap tube feeds into the evaporator, which is the heatsink that sits on your processor. Inside the evaporator, the liquefied gas encounters an open, lower pressure region and flashes back to a gas. This phase change is an endothermic reaction, which means the change of phase sucks up heat, and a lot of it at that. This heat laden gas is then sucked back into the compressor via the suction line, but even though it has picked up heat from the evaporator, the gas is still in the region of -15 to -60 Celsius, so when the compressor sucks it up, the compressor is partially cooled by the gas. Now if this paragraph didn’t make any sense at all, I suggest grabbing a cup of coffee and read it over again and look at the diagram.


To make one of these units, you will be needing the following materials. While you may need a few more things or not everything that I listed here, in general this list should work.

1) Compressor larger than 1/8th Horse Power
2) Decent sized condenser, try and match it to your compressor, if it was attached to the compressor, you are set.
3) Capillary Tubing, preferably .7 millimeters
4) 1/4″ Inner Diameter (I.D.) copper tube
5) Suction line material, I prefer to use a 3/8″ I.D. copper tubing for my suction lines, but some people use flexible lines, especially flexible stainless steel lines. Just remember it has to be able to hold pressure, preferably 400+ PSI.
6) An evaporator, it can be anything, so long as it has a surface to place against your processor, and enough surface area inside to let the gas to its work.
7) 2-3 Schrader valves. These valves are sort of like the valves that are on a pneumatic tire, threaded with a pin valve. These can be found at an HVAC store, usually brazed to a short section of copper pipe.
8) 2-3 Copper T sections, these sections are used to attach the valves to your system.
9) 1 filter/dryer, this specialized piece of pipe has a chemical in it that helps remove any moisture and debris
10) Insulating materials, preferably insulating tape and a short length of Armaflex tubing.
11) Tools, to be discussed below

Once you have sourced all of these parts, you are set. Just a warning before you go about tearing into home appliances to find compressors, in the United States there are laws and fines that don’t permit you to vent refrigerants into the atmosphere. If you decide to use a compressor and condenser from a working appliance such as a dehumidifier or refrigerator, first take the unit to a HVAC store and have the suction all the refrigerants out. If you are lucky they will do it for free, otherwise it won’t cost you much at all. The laws that I mentioned above are for any regulated refrigerants, so your choice of gas will be very limited, but I will get into that later.


To make all those materials into a useful product, you will need a lengthy list of tools. The tools are the major barrier to entry in this field, these tools will cost you $200 to $300 for the cheapest set and even then you have to improvise a few things. I won’t be listing individual prices of tools because the prices may fluctuate depending on your country/region.

1) A brazing torch, this torch is used to heat up the metals and braze them together. Brazing will be discussed later.
2) Brazing rods, these rods are the material that you connect pipes together with.
3) Flaring tools, this clamp and tool let you “flare” a pipe, making it easier to put pipes together.
4) Pipe cutters, this tool lets you make clean cuts on pipes without bending and crimping the pipe

5) Charging manifold and hose, this manifold lets you connect, with the hoses, your system to your bottle of refrigerant. Usually these manifolds also feature gauges to show you the pressures on the various lines.
6) Vacuum pump, in order to operate correctly, your system must be vacuumed to remove all air and moisture from the lines. Professional vacuum pumps can cost a decent amount and provide you with the best vacuum, however I use a 1/3 horse power rotary compressor as my vacuum pump. Compressors can be used as vacuum pumps but they won’t perform as well as a profession vacuum pump.
7) Temperature probe, you need a probe to determine the temperature of your evaporator, sticking your finger on it will give you a quick case of frostbite.

Once you have rounded up all of these tools and materials, you can begin assembling your setup. To start, you should practice brazing copper together and once you can do that properly, you can begin putting stuff together.

Brazing 101

Brazing is the name for the process of using a copper/silver rod to melt two copper pipes together. Brazing is the only way you can get a decent connection that will hold the pressures required. Some people use solder to connect pipes together, but for me solder doesn’t form a strong enough bond, and solder has a higher chance of cracking and breaking. In order to braze copper together, a few things need to come together at the same time and then it is smooth sailing. Here is a quick rundown on how to braze.

1) Designate one piece as the female and one piece as the male. The female piece should be flared out with your flaring tool so that the male will fit snugly within the flared end of the female.
2) Heat the pieces of copper till they emit a faint cherry colored glow. The hotter you get the copper, the easier it’ll be to braze the pieces, however you do not want to literally melt the copper you are trying to connect.
3) Once both pieces are glowing cherry red, push them together in the shape you want them to hold and make sure they fit snug.
4) Lightly touch the connection with the brazing rod while applying heat to the connection. If you are at the right temperature, once the rod heats up it should start to flow like warm butter around your connection. This will happen very quickly, so do not be surprised. Make sure there are no gaps; don’t worry about it being pretty right now, this is just practice.
5) Let the piece slowly cool. I know it might be cool to quench your new connection in water, but all this will do is cause unneeded stress and it introduces the possibility of cracking your connection. Remember, you aren’t a blacksmith, you are a brazing phase cooling fiend.
6) Once cool, rub the connection clean with an abrasive material, I use steel wool. Once the connection is clean, inspect it, looking for cracks, pinholes, and defective workmanship. The first couple of times you are guaranteed to screw up, it just takes practice.

Note: When you heat up copper in the presence of oxygen, it will oxidize, causing that layer of black soot to form on the pipes. The insides of the pipes aren’t exempt from this either, so your best bet would be to use a shielding gas. Shielding gases are non-oxidizing gases that are pushed through the piece you are brazing to keep the copper clean. I only use a shielding gas when I am brazing the evaporator and components near the evaporator, everything else can usually be blown out with compressed air.

There you have it, the most important skill to this project is your ability to braze. Now we need to figure out how to put all of those parts together so that they make sense and are functional. On your first try, assembling one of these could take upwards of 20-30 hours if you work slowly but carefully, remember safety first, you are dealing with compressed gases, which aren’t happy gases.


Find a clear space outside that has access to electricity and a roof if you decide to build in the rain. Layout all your pieces and go over the above checklist, it is terrible to be missing a piece or tool when you are at a crucial stage. Stock up on your brazing and refrigerant gases, you’ll go through them quickly, especially the brazing gas. Once you have everything in place, start by placing your compressor on a platform and build your unit around it. There are many techniques for building these units and many variations, so I will only the two most common.

1) Cap tube inside of suction line. This design places the cap tube inside of the suction line, so there is only 1 piece of copper connected to your evap. This design increases safety by protecting the high pressure cap tube, but it makes connection your evaporator more complex due to the integrated cap tube.
2) Spiral wound cap tube. This design spirals the cap tube around the exterior of the suction line. In this case you should gain better sub cooling by cooling the cap tube with the cool suction line. While you will get sub cooling in the first design, only a few feet will be exposed to the gas, while the spiral wound exposes the entire line to cool temperatures.

These designs differ on in the positioning of the cap tube. I prefer to use the first design, while it makes brazing up the system a bit more complicated, it adds a bit of safety which I feel is an appropriate trade. Make your choice and stick with it though, once you start you can’t go back.

Cap Tube length

The length of your cap tube will be determined by many things. First off, you need to figure out exactly how much heat your processor will be putting out. I always err on the side of caution and calculate for 20% higher than what I expect to happen. The formula for processor output is the following.

TDP_new = TDP * (MHz_new/MHz_stock) * (V_new/V_stock)^2

If you want an explanation as to what this really means, view my article Overclocking Series: Basics which explains everything. Now working with captube, the equation is use works as the following.

The base measurement that you will work with will be 10 feet of .028″ diameter with r404A. This base will handle a load of ~150 watts of heat. For every 15 watts extra you will cut a foot and for 15 less you will add a foot. Once you have the proper length for your heat output, you will adjust for your gas. For r290 and r22 you will shorten by a foot. For r12 you will shorten by 3 feet. For r134a you will shorten by 4 feet. Now that you have your length, you can adjust for diameter. If you are using .028″ tube, you are done, however if you are using .026″ tube, multiply by .7. If you are using .031″ tube, multiply by 1.6. Here is an example to explain everything.

180 Watts, using .031" and r290.
180 - 150 = 30 watts excess. 30 watts = shorten by 2 feet.
So we have 8 feet. For r290 you shorten an additional foot, so 7 feet. Now with .031" tube, you multiply by 1.6, so the final length is 11.2 feet of .031" cap tube for 180 watt load.

Evaporator Design

Of all the critical pieces in your systems that will affect your performance, this piece is critical. There are a few requirements to make this thing work. First off, the surface must be smooth and have enough contact area to touch your processors entire integrated heat spreader. Second off, you must have enough surface area inside the evaporator so that the gas can pull off as much heat as possible. Just like in water cooling, the more surface area, the more effective your system will be. The last requirement is structural integrity. Remember, this piece will be exposed to very low temperatures while operating with very high pressure gases. I don’t recommend using aluminum as a material for your evaporator, silver can be used but is very expensive, so copper is your best bet. If you can manage all of these requirements, your evaporator should do just fine. If you do not have access to a lathe or CNC machine, I highly recommend going to a plumbing store and buying a few copper end caps and brazing them together. While it won’t have as much mass as a custom made evap, it is easy to build, cheap, and easy to replace. I personally use 3 end caps, a 2″, 1.5″ and 1″. With the 2″ facing up, I place the 1.5″ and 1″ inside the 2″ facing down and braze them into place, drilling holes into the 1″ and a single hole into the 1.5″ to permit the gas to escape. My suction line connects to the single hole on the 1.5″ cap and my cap tube is feed through this same hole. The complexity of your evaporator will depend on your access to machinery and tools, but even the simplest evaps will at least net you negative temperatures.

Notes on construction

When brazing your evaporator up, you will want to minimize the interiors contact with oxygen. When I braze my evaps together, I put everything together, unbrazed, and then I connect my cap tube to a propane cylinder and leak in 2-3 PSI of propane. This propane will flow through the cap tube, into the evaporator, and then up and out the suction tube. At the end of the suction tube I ignite the propane and a small flame will flicker. This technique will shield all portions of your evap and suction line, however the danger from propane is immense if you do not do this correctly. I do not recommend this technique to anyone unfamiliar with propane, I only use this because it makes my life less complicated, most other people use dry nitrogen gas to shield with. Nitrogen is ideal because it is inert, relatively cheap, and non-toxic.

You should attach a T connection on the high and low pressure sides of your system. The high side is the portion between the compressor and the evaporator, and the low side is the suction line. I place a T connection immediately after the compressor and I place another T connection where I feed in my cap tube into the suction line. On the end of each T, you should place a Schrader valve so that you can charge your system and monitor the high and low pressures.

If you are using a solid suction line, you will want to make a coil with your suction line to reduce the vibrations. Compressors produce vibrations on all the lines, and if your evaporator vibrates too much, there is a risk to cracking your core. I place my coil between the T connection and the evap, preferably right before the T connection. I personally loop the copper 3-4 times in a 4″ diameter coil and I have found this to be very effective at reducing all but the most extreme vibrations.

When positioning your condenser, you will want to make sure the gas flows from top to bottom. The inlet should be near the top row and the outlet near the bottom, otherwise if any gas starts to liquefy within the evaporator, you’ll end up with a puddle of fluid at the bottom and it will require a heavy charge to get any performance. In regards to placement, ideally you want it 6-10 inches away from the compressor with a fan between them. By placing a shroud over the fan and condenser, while having the fan suck air through the condenser, you will also end up blowing air over the compressor, saving yourself the trouble of worrying about cooling the compressor. Otherwise, you can put them close together and shroud the condenser and blow air through the condenser, but people have found this to be slightly less effective.

Insulate your entire suction line and any part that will be below ambient. My entire suction line is insulated up till 2″ away from the compressor, you do not want the lines exposed to the environment or you will end up cooling a lot of air also. If your unit is not insulated and you are seeing terrible temperatures, once you insulate you will probably be surprised with what your unit is capable of doing.

Gas Selection

In the United States, you must have a HVAC license to be able to purchase any commercial refrigerants. For those that do not have a license, you are limited to the gases that you can use. Here is a list of the various single stage gases you can use without a license. I will not be listing all the various refrigerants that can be obtained with a license, if you have a license, you will know what you can get, if you don’t have a license, get one first, then worry about those gases.

Propane- the poor mans gas of choice. This wonderful hydrocarbon is great at roasting steaks, heating water, and cooling computers. Propane is cheap, plentiful, relatively low pressure, and did I mention cheap? Now the downsides, the boiling point isn’t that great, a measly -42.09 Celsius, and it is flammable when exposed to oxygen. Propane is also heavier than air, so if it does leak in a room, it will sink to the floor where it will puddle and form a thin layer of propane gas. Propane naturally has no smell, but the chemical ethanethiol is added to propane to provide it with a distinct smell.
R-134a- this gas is what has replaced R-12 in the refrigerant industry due to its non-ozone depleting properties. This gas can be obtained from various automotive recharging kits and in some rare cases from HVAC stores. R-134a has a very high boiling point of -26.08 Celsius but it condenses very easily and it is non toxic. One note about safety, R-134a is denser than air and it will displace oxygen in lungs, potentially causing asphyxiation if inhaled in large quantities.
CO2- carbon dioxide, while commonly known for its solid form, dry ice, can be liquefied and used as a refrigerant. CO2 has a boiling point of -78 Celsius and it can be added in small quantities to a r290 charged system to drop temperatures. Other types of phase coolers can use CO2 as the primary gas, but that is beyond this guide. CO2 is toxic if inhaled in large quantities but it is relatively cheap and easy to purchase, many Paintball fields can sell you liquid CO2.

There are other gases that you can obtain, but it is either illegal or a complex process to be able to purchase those gases, so I have stuck with these three gases. Once you are very familiar with refrigeration, you can advance to higher pressure gases and lower temperatures. I have had r290 charged systems down to -60 Celsius by pulling a high vacuum, so it is possible to get decent temperatures with these unregulated gases.


When you brazed up your evap, the exterior will have oxidized and it will have roughed up the surface, no matter how careful you are. With the system completely discharged, you will want to use some sandpaper and a flat object to grind your evaporator flat. Gradually work you way with 180, 220, 400, 800 and even 1200 grit if you can get it. The flatter your evap, the better contact you will achieve and the better your temperatures will be. If you do not lap it, your contact patch will be uneven and small, and your temperatures will spike under load.

Testing and charging

Once you have put your entire unit together, you must test its integrity before you can put it under load. I use propane as my purging, charging, and testing gas just to make my life simple, but you can use any “dry” gas. You do not want to use any gas that has any moisture in it, such as compressed air, you are trying to keep water out, not put water in. On the first charge, I pressurize the system to 25 PSI and then test for leaks. With propane, your nose is a great piece of equipment to test if there is a leak, if you even smell propane, you have a leak. The best way to test for leaks involves getting some dish soap and water, mixing it well and then slathering this bubbly concoction over every joint on your system. If you have a leak, bubbles will start to form immediately. Wipe off the soap and mark every place that there was a leak, discharge the system, and patch the leaks with your brazing rods. You want this system to be air tight, even a small leak can cause problems.

Once you are certain that your system has no leaks, begin to slowly pressurize your system, slowly raising the pressure. When I do this, I put my unit behind a low wall of bricks and I sit behind the wall with my pressure gauges, and I slowly charge the system, safely. I charge the system till around 350-450 PSI and then let it sit for 24 hours, exposed to the sun. The heat from the sun will cause the gas to expand further to 500+ PSI, and if your unit holds that for 24 hours, you know it is solid. Ideally you will charge it at one time, record the pressures once the gas settles, and then check back a day later, they shouldn’t fluctuate much, if any at all. Just make sure when you check the pressures that the ambient temperatures are about equal, otherwise the gas will have expanded to different pressures.

Once you are sure your unit can hold pressure, you will begin to vacuum and purge it. For the first run, hook up your vacuum pump and suck all gas out of the system till you are running a high vacuum. Then get a hairdryer and slowly warm up every component on your system, starting from the compressor and working your way around the loop. By heating up your system, you help release any moisture in the system and hopefully you will remove it from the system. After 3 to 6 hours, charge your system to 50-100 PSI, and repeat the process. This may seem like a waste of gas, but this will help ensure that you are removing all moisture from the lines, moisture is the enemy. Do this 2 to 3 times, and on the last run pull a vacuum for 12 or more hours. This may seem excessive, but this will give you the best results. When you are finished vacuuming, make sure you never open the system up to the atmosphere, you should begin charging with your refrigerant, charge your lines and make sure the lines have no air in them. With propane, I like to vent a small quantity of propane from the lines as I connect them, making sure there is only propane in the lines. Once you have everything hooked up, slowly add your gas till you reach somewhere around 75 PSI. Now you begin tuning.


Depending on what results you are looking for, the amount of gas you use in your system will vary. Tuning a Single Stage single gas system is rather easy, the more gas you have in the system, the higher your temperatures will be but you will carry more load. For example, on my system when it is operating, if the low-side has between 5 and 10 PSI of pressure, my temperatures hover around -10* Celsius. If I remove some gas and drop the pressure to 0 PSI on the low-side, my temperatures will drop to around -25* Celsius. If I continue to remove gas and start running vacuum on the low-side, my temperatures will continue to drop till they max out around -60* Celsius with -22″ Hg. A quick side note, vacuum is measured in inches of Mercury, which most pressure gauges can read to a certain degree of accuracy. In tuning your system, you should apply the correct amount of heat that you are tuning for, meaning you either need a dummy load of transistors to generate heat, or you need to attach your system to your processor while doing the final stage tuning. Since I am running on an Athlon 64 and they don’t like extreme temperatures, I have tuned my system to hold the temperatures around -25* Celsius at 2.95GHz and around 1.5 volts. You just need to play with it, but just remember, the more gas, the higher the temperature but the more heat it can carry. If you are tuning for a water chiller, remember that unless you are running a high anti-freeze/water ratio, you will want to keep your temperatures around zero, so tune for a load around zero. Here is an image of my pressures under operation. One quick note on charging, you will want to add gas via the low-side, preferably while the unit is in operation so it will suck the gas in. Discharging gas can either be done with the unit off or on, but you will want to discharge from the high-side to prevent air from being sucked into the unit.

Mounting and Insulating

edit: This guide is slowly being reworked and split up into multiple articles to let me expand upon the concepts in a more clear fashion. An updated guide for the LGA775 socket has been posted here, the guide below is based on s939 for AMD and uses less-refined techniques for insulation. I’d highly recommend checking out the LGA775 guide even if you have an AMD or s478 motherboard.

Ideally you want to produce enough mounting pressure to securely fasten your evaporator to you socket, but you don’t want to over do it and crush your core. For my mounting bracket I used two pieces of steel and two threaded steel rods and made a sandwich with the motherboard between the pieces of steel. You can create whatever system will work for your motherboard, just make sure you have enough clamping pressure to make good contact. Once you have your mounting system created, you must begin insulating. The principle behind insulation for electronics is to form a perfect seal. If no moisture can enter your socket, there will be no condensation, and that means no short circuits. To prep your board you will first want to use some form of lacquer to seal the PCB and the surrounding electronic bits and pieces to protect them from any condensation that may form. I myself used Revlon clear nail polish because it dries quickly, is easy to spread, easy to remove, and best of all, doesn’t look too funny. I try to insulate a good 2-3 inches from the center of the socket just to make sure no potential condensation has a chance of setting in and causing any damage. Make sure you insulate both the front and the back, just because you don’t see the backside of the board doesn’t mean condensation won’t form.

Once you have coated the area around the socket with a sealant, you will begin applying foam insulation. Foam insulation is effective because it keeps moist air from coming in contact with your very cold processor. The whole purpose of the insulation is to prevent your motherboard from sweating and shorting itself out. The main rule in insulating is to prevent any air contact with the cold pieces, so try to overlap everything and seal everything very well. My choice for sealing the motherboard was Armacells Insulating tape. This stuff is 2 inches wide and an eighth of an inch thick, and best of all, it came in a 30 foot roll. To start I worked on the back, cutting 6 inch strips and covering a 6″ by 6″ square on the back of the board, covering the backside of the socket. Then I cut 3 more strips and laid then across the previous strips, sealing up all the gaps. Then I placed another layer of tape, giving me a 3/8″ of insulation on the back of the board. Even with this much insulation, the foam would get a bit cold but no condensation would form. Now to work on the front.

The front of the board is a bit more complicated due to all the little surface mounted bits and pieces. Here I decided to insulate everything between my ram slots, PATA sockets, power regulators, and AGP slot. There weren’t many capacitors and mosfets in this area, so I didn’t have to trim my insulation too much, but if you look at the picture above, you can see how I trimmed the insulation to fit around snuggly around the capacitors. Just like the back of the board, you will want to get 2-3 layers of insulation down, while maintaining a tight seal around the socket. Do not plan on removing the CPU once you have sealed up the socket, otherwise you will ruin your seal. In regards to the socket itself, some people pack di-electric grease into the pins to keep any air out of them, I didn’t want to dirty up my socket so I skipped that step. Just remember that this insulation is what keeps your motherboard alive, so if you feel like there isn’t enough, add some more. Once your motherboard is all sealed up, you need to insulate the evaporator and suction line. Since the evaporator gets much colder than the surrounding motherboard, your insulation will be much thicker here. I ended up with nearly an inch of insulation covering the evaporator, I suggest putting the evaporator onto your processor with your computer off, and fire up the compressor. Let everything get down to the running temperature and then start feeling for seeping cold. If you find any, slap some insulation on it. Eventually you’ll get to a point where you won’t feel any cold seeping out, then add another layer for safety and that part is done.

For your suction line, you should insulate it so that it won’t sweat, if it sweats the water could leak down it and spread over the evap, ruining your insulation there. I put a layer of insulating tape over the entire suction line and then put a 3/8″ thick Armaflex tube over the suction line. The rubber foam on the Armaflex really kept the cold in, that has worked for me. Do the same for everything that gets cold on your system, humidity is the enemy here.

That about wraps this article up, I will be building an autocascade sometime this summer and I will try to do a better job of recording everything on film and making a detailed guide to that too. If you need any clarification, please let me know.

Some results

Earlier yesterday I had enough time to do a quick 30 minute bench session during which I started running SuperPi 1M until the processor was unstable. I pushed the voltage to ~1.7 volts, stock being 1.4 volts, and I dropped the Multiplier from x10 to x8 and then started to push the clock speeds. Computer crashed at 324×8, up from a stock 200×10, so I think around 320 my memory controller is unstable. If only AMD still had an off-chip memory controller, this processor could really scream. Here are some of the results, later today I will do another bench run and update this a little bit, still a work in progress and it will probably never be done.

If you enjoyed this, please Digg me here.

There will be more results to come, let me know if something doesn’t work or if you need a more detailed explanation of this. In regards to actually building one of these, I highly suggest you read up more than just this guide, this is just starting material, still lots of other little things you need to know. The following links will really help you gain a complete understanding.

Basic Overclocking
Memory Overclocking
Introduction to Cooling Cooling Section Phase Change Guides
Team Puss


Finished my round of benchmarking with SuperPi 1M. The highest stable overclock was 2860MHz, while the highest unstable overclock was a 3218MHz. I didn’t manage to snap a picture of the 3218MHz due to whenever I would copy the screenshot and then past it into MSPaint, the computer would crash when saving it to the hard drive. Here are the stats on the hardware I used, and I will post up my 3DMark 01,03,05 and 06, along with CPUMark0X and any other benchies I try.

Processor: AMD Athlon 64 3200+ Winchester 200MHz x 10 @ 1.4 volts
Motherboard: MSI Neo2 Platinum
Memory: 2x 512mb OCZ Platinum Rev2
Video Card: Nvidia mx440 AGP and ATi X700 256mb Pro AGP
Powersupply: Ultra x-connect 500 watt w/ tacky green neons
Hard drive: TriGem 4Gb 5400RPM PATA

The Buzz {16 trackbacks/pingbacks}

  1. Trackback: on April 17, 2006
  2. Pingback: Blah, Blah, Blahg » Build your own phase-change cooling system on April 19, 2006
  3. Pingback: Tech Growth at on April 19, 2006
  4. Pingback: GPI and Co » Blog Archive » Geek’s Powa on May 14, 2006
  5. Trackback: MAKE: Blog on May 15, 2006
  6. Pingback: Daily DIY » Blog Archive » Vapor Phase Change Cooling on May 16, 2006
  7. Pingback: Velcro City Tourist Board » Blog Archive » Links for 17-05-2006 on May 16, 2006
  8. Pingback: File Compression Review at on July 8, 2006
  9. Pingback: Introduction to Cascades and Autocascades at on February 9, 2007
  10. Pingback: Put ice in foil - Page 7 - Computer Forums on April 8, 2007
  11. Pingback: Phase Change LGA775 Socket Insulation Guide at on May 14, 2007
  12. Pingback: Case Modding For Dummies on March 21, 2008
  13. Pingback: Top Guild Computer System Specs - st0ned Guild on May 14, 2008
  14. Pingback: Overheard: Chriss Morell explains phase change cooling - Overheard in the tech blogosphere on November 20, 2008
  15. Trackback: Sammy Sullivan on June 30, 2009
  16. Pingback: Window Unit AC's --> CPU Cooler? - - on April 6, 2010

The Conversation {56 comments}

  1. Paul Stamatiou {Monday April 17, 2006 @ 10:52 pm}

    In case you didn’t get my text msg at 2:30am…

    U r on
    ull be on makezine later today, he emailed me

    goot wurk ***** nugget

  2. Grimiore {Tuesday April 18, 2006 @ 12:27 am}


  3. Luke Labern {Tuesday April 18, 2006 @ 5:42 am}

    Awesomely thorough.

  4. Chris Morrell {Tuesday April 18, 2006 @ 6:27 am}

    Thanks for the comments everyone, I will be benching this system again to try and get 3GHz SuperPi stable, but this chip might be reaching its limit. Hope you enjoyed the detailed read.

  5. Meagan {Tuesday April 18, 2006 @ 6:33 am}

    Ahhhh, I see. This is where you’ve been everyday for the past I don’t know how long. Congratulations! It’s really cool, Chris. You should make me one ;)

  6. x {Tuesday April 18, 2006 @ 9:18 am}

    very nice, good work

  7. chris ;) {Tuesday April 18, 2006 @ 9:45 am}

    nice work man ;) also this project isnt that what i understand under mechanical engineering it s straigt awesome. and con gratz for the hack a day ;)

  8. Your Dad {Tuesday April 18, 2006 @ 10:12 am}

    This is fantastic, your one in a million….Your Mom and I love you very much…..

  9. Dale White {Tuesday April 18, 2006 @ 10:21 am}

    Hey, I recognize some of those parts from as being stolen from the Battle Lab …….. Very impressive Chris.

  10. Chris Morrell {Tuesday April 18, 2006 @ 10:38 am}

    Ahh! That is where that thermometer came from. I found it in a box a month ago and was wondering where it had come from, I have so many random bits in pieces in this room. It was attached to my first single stage unit that I built over the summer while at the Lab. If you guys need it back, I can send it back, it has a few scratches but they add character, and it has a few wild stories to tell. Fedex can have it there in a day, just say the word.

    Off topic, just hit 3287 MHz and computer crashed, that is the highest I’ll be going, no need to risk burning this processor out.

  11. Meagan {Tuesday April 18, 2006 @ 11:20 am}

    Wow, Chris. You stole from the military. That’s not good.

  12. Albert {Tuesday April 18, 2006 @ 12:42 pm}

    Simply ROCKS!!! great page!

  13. Stego {Tuesday April 18, 2006 @ 2:30 pm}

    Nice, great Pc and work, congratulations to the author(s)


  14. Circle {Tuesday April 18, 2006 @ 3:01 pm}

    Hey man, I bet it’s your ram, cause in that last screenie you’re running the ram at 250 (DDR 500) I know that it’s a pain but try dropping the memory divisor to 1/2.

    Winchester at those speeds, very impressive. Winny’s are not good overclockers, and being that you have one a 3ghz is very impressive. After building a system like that I might consider a Venice core chip, they hit higher speeds.

    I don’t know about the capacity of your system, but if you have a lot of room for more thermals and are feeling a bit radical you might condier adding a peltier, but I’m not sure of your current output (from that screaming winny) or if you could even find a peltier rated high enough. (Of course there is also the overhead generated by a peltier to compensate for.)

    I’ve done a few peltier cooled rigs and they’re a walk in the park compared to this. (but I would NEVER be able to hit those temps without a ridiculous cascading system, which would at the bare minimum require a secondary PSU of over a KILOWATT. LOL.) Insane props man, Incredible skill, incredible throughness.

    Also around those reference clocks many boards become unstable. I know mine holds to 350 or so, but most poop out around 300 - 320. Something to think about (cause a 1 mhz error there = 10mhz at the proc.) Try the memory divisor, that will eliminate the possibility of the ram or the mem controller being your problem.

    Keep embarassing my systems like this and I won’t even show up at a LAN party with it anymore! LOL. Good work.

  15. plbg32 {Tuesday April 18, 2006 @ 3:49 pm}

    very cool work you’ve done there, this is a project i’ve wanted to do for a long time, but never got around to it. anyhow the point being i’m a plumber by trade and a place to get the parts for your rig like this if you are so inclined is an old drinking fountain with a cooler. its already set up and the conversion may not even need much retofitting and regassing of freon.
    once again great job!!!!

  16. Geoff {Tuesday April 18, 2006 @ 5:45 pm}

    Nice work, now try building a 3-stage cascade that holds -125C :P

  17. Tom Johnston {Tuesday April 18, 2006 @ 8:01 pm}

    Awesome implimentation. As I’m sure you know, petrol refineries have been using propane as a refrigerant for a long time (can’t remember it’s “r” designation). Be careful with that shit! I can only imagine the trouble an over energetic, sloppy geek could get into trying this. Darwin awards may occur. LOL. So be it!

    Hey, on a legal note… getting the proper certs to legally buy and use real “freon” is a pretty easy. Any HVAC supplier can tell you where you can take the EPA Section 608 test. This test is pretty easy to pass. It’s mostly about safety (common sense at that). Having that cert opens the door to refers that are fine tuned to certain temp ranges. Too bad you have to buy those refers in 20 pound lots. Doh! R134a is a nice certless starting point for working with traditional refers. I (we) prefer R416 for R12/R134a apps. It solves some probs that R134a has (oil migration, varnishing, ect.) Gotta have the cert though.

    Keep up the good work. This has inspired me to merge my trade with my hobby.

    Tom Johnston
    Idaho HVAC Journeyman #1251

  18. XD {Tuesday April 18, 2006 @ 8:22 pm}

    hey my names Steve, im thinking about doin a similar project with an 8 CPU Xeon box and if you have any hints you didnt include on here id like to have them before i go ahead with this.. make sure you put word “overclock Xeon” in subject..i think that e-mail box has like 3000 emails curently lol also props to ya for great work

  19. todd@hvactalk {Tuesday April 18, 2006 @ 9:36 pm}

    great description, and nice job setting that up. its good to see you really did your background work before trying this out and took condensation and all into account. very good technical description of the refrigeration side layout. all in all very impressive.

  20. P {Thursday April 20, 2006 @ 9:26 am}

    And, what’s the point??? Cool work but lame concept.

  21. Chris Morrell {Thursday April 20, 2006 @ 3:01 pm}

    Well if you look at the last couple of pictures, you will notice I have pictures of a 50% overclock. I ended up hitting 3287MHz and crashing Windows XP, the memory controller was actually too cold for the HTT speeds. Besides the performance, it is fun to see what your chip can do, compare this to dyno testing cars, there is no point to spinning up huge drums with your car, except to understand how it performs. It isn’t for everyone, and not everyone will understand, thankfully.

  22. Larz {Thursday April 20, 2006 @ 3:32 pm}

    Great work!

    Can you post a picture of your evaporator made from stock end caps?


  23. Psyqwix {Thursday April 20, 2006 @ 4:16 pm}

    I’ve never really been into engineering myself, but I must admit this article is intriguing! Loads of detail and really interesting, well done!

  24. Silvia {Saturday April 22, 2006 @ 3:26 pm}

    This is fantastic, you have inspired several people and how great is that?!!!!!!! Good for you, we are so proud of you.

  25. Sturla Røysland {Sunday April 23, 2006 @ 1:59 pm}

    Dude! I thought this was an old guide, and that any chance of asking some questions would be in vain.

    Omg, ILU! Lardases guides just went to hell, as the pictures from the ftp was removed. I can’t tell you how relieved I was when I stumbled over this EXCELLENT guide. If I could I would’ve kissed you. Thanks for cheering up sadfaced Norwegian!

    But, I think you went a little too fast at the end there. If it’s not too much to ask for, could you possibly add some extra material on the whole insulating ordeal? And also a little extra on the charging bit? I’m not really sure of where to “put in” the gas, and which of the lines from the charging-unit to use.

    But all in all, tnx for making me take yet another leap into the world of phase-change-cooling :D:D:D:D:D:D

  26. coldmizer {Thursday May 4, 2006 @ 5:47 pm}

    Great site!!!! Found your site because I want to build one myself too. Couple questions: How much did it cost? How noisy is it?
    Thanks. Good job.

  27. Chris Morrell {Thursday May 4, 2006 @ 7:48 pm}

    I can’t answer your questions if I haven’t got an email address to reply to. Depending on the size of the unit and how you insulate the noise can be unbearable or moderatly loud, but it won’t be silent. A lot factors into the noise production of a unit including compressor, mounting, and insulation. Costs are variable depending on your connections and what you can build rather than buy, for me it ran about $300 for my tools and first two units.

  28. coldmizer {Friday May 5, 2006 @ 1:22 pm}

    Great. $300 is quite a bit less than a Vapochill (where I first learned of phase chance cooling).
    How difficult do you think it would be to build a system with additional coolers for the GPU and RAM,
    or would this be overkill?

  29. Selki {Saturday May 6, 2006 @ 12:53 pm}

    Now I have to go try this out! Wonderfull guide. I am also interested in additional coolers for the GPU. How difficult would it be to add a GPU cooler onto your build and should a 1/5HP Compressor work ok?

  30. The Visionary {Wednesday May 17, 2006 @ 11:50 pm}

    Could you post some links to supplies on the internet… Like compressors and such. I have plans for this shtuff… =+)

  31. Catalyst {Sunday May 28, 2006 @ 8:14 pm}

    just an idea not as affective but probably easier….just wondering how well it would work…if you took a mini fridge apart and make a tank for water cooling….build the tank around coils of the fridge…and just continued as normal for water cooling…what do you think estimated tempatures would be…would love feedback


  32. Chris Morrell {Monday May 29, 2006 @ 2:44 pm}

    Many people actually use water fountains for chilled water coolers instead of going through all the trouble to create a direct die system like I have created. The only issue with chilled water is that you have to insulate every hose and then have a separate container for water, pump, fans, and radiator. You can hit temperatures around -10* Celsius if you insulate well and even lower depending on the effectiveness of your heat exchanger and the compressor itself. Let me know if you have any other questions.

  33. ZeroCool34 {Friday June 9, 2006 @ 5:11 pm}

    Great job man! check out what our school did!


  34. Chris Morrell {Friday June 9, 2006 @ 5:31 pm}

    That is a cool project, I always wanted to get a Pentium 4 and beat the hell out of it. I’d prefer to get a triple digit autocascade instead of LN2, I would much rather have 24/7 performance. Very cool project, that had to be fun setting up and executing it. Did the board work afterwards, noticed a large amount of frost on the capacitors.

  35. Kevin {Friday June 9, 2006 @ 8:37 pm}

    Hey I am building my own phase change system. What did you use to control the evap temp and the compressor cycles. My unit is going to be in a custom built case. It is nice to know that there is someone else out there like me. thanks alot. oh and thanks for the calculations. God Bless.

  36. Tsepo {Thursday June 22, 2006 @ 2:50 pm}

    this is the kind’a thing I’ve been waiting 2 c, great development chris, shld take it further, cos the heat sink on the intel processors sucks, just feels the system with extra heat which ends up increasing the system temp…I tried sealing the open holes except for one and sucking the air out wit a vacuum cleaner like device, it kind’a worked but I guess I need to do a bit more work on it to reach a certain level…

    great work…frm tsepo

  37. tyler {Friday July 7, 2006 @ 11:01 pm}

    I did this awhile ago when I was younger with an old mini fridge and a tandy

  38. Andrew {Sunday July 9, 2006 @ 10:14 am}

    Hey ZeroCool34,

    Before you steal someone elses work, make sure you get read of their name in the credits.
    That video and project was from Toms Hardware along-time ago.


  39. Xone God {Tuesday August 1, 2006 @ 9:00 pm}

    Its humbling to know that there are people other willing to thinker and create from their minds to the work table. It’s a true relief to see there are more of us out there. I too am building two similar units one for testing other available liquid gasses and one for my servers. I applaud you.

    Xone God

  40. Tzvier {Saturday August 5, 2006 @ 4:21 am}

    Well done. Probably very soon I will be following this guide quite closely. I just wanted to give a little tid-bit back. On the topic of lapping, toothpaste can be used to give a glass-like finish. I believe it is similar to 2000ish grit sandpaper.

  41. Chris Morrell {Saturday August 5, 2006 @ 8:35 am}

    I never thought of using toothpaste as a lapping material but after thinking about it I suppose that would do the trick as a final polish. Nice idea

  42. Rick Johnson {Sunday October 1, 2006 @ 5:44 am}

    When you vacuum the compressor down, doesn’t it suck out all of the oil? I know that there are 3 tubes coming out of the compressor, 1 for suction, 1 for pressure and the other 1 is pinched off. I always figured that the pinched off 1 was for putting in the charge. If I am wrong would the pinched off 1 be for putting in the oil?
    You mentioned refrigerants R12, R134a, Propane and CO2 as the charge, have you ever tried Freeze 12? It’s like R134a, you don’t need a license to get it and it is ozone safe?

  43. Chris Morrell {Sunday October 1, 2006 @ 9:07 am}

    I have never used the service port to vacuum down a system due to the close proximity to the compressor and as you mentioned, compressor oil. I always install 2 ports in the system, one between the compressor and the condenser and one between the evaporator and the compressor, so that I have the ability to monitor the pressures and to vacuum the system without fear of sucking out too much oil. I have heard of Freeze 12 however I have never had any experience with it. I would be concerned about its ability to hold a charge but I never used R-12 and I don’t know much about R-12’s physical properties. Just remember that the boiling point of the liquid will be the limiting temperature in your system and you must avoid operating your system around the triple point so that you don’t end up with the solid form of your refrigerant(this problem is common in CO2/R290 autocascades). It does look like an interesting gas though, if I can find a local supply at my gas depot I might pick up a cylinder.

  44. Paul {Monday December 4, 2006 @ 6:20 pm}

    Hey Chris could you e-mail me. I would like to talk to you more about your project. I’m in the same business and have a great solution for mounting the evaporator head to any intel clip.

  45. brad johnson {Friday February 23, 2007 @ 12:39 pm}

    i just need to pick your brain i have taken the cooling system out of a water cooler i just need to know if i need to do any work to it to make it work other then makeing a evap to mount to my cpu. i have some pics if that will help

  46. Ryan Medina {Monday April 16, 2007 @ 12:00 pm}

    Nice Guide , im trying this myself over the summer and have read over this a few times now :)

    Thanks alot man :D

  47. nickaname {Monday June 25, 2007 @ 5:57 am}

    hey aehm,… some of your stuff sounds a bit like plagiarism. seen the captube length estimations in a post from 2004 on the xtremesystem site (yours is a 2006 project, right?). at least a reference should be made when using exactly the same words or at least make it look like a quote…???

  48. Chris Morrell {Monday June 25, 2007 @ 11:25 am}

    I link back to XtremeSytems quite a bit and I also feature all of my builds on XS. Gary Lloyd’s captube formula is a publicly available formula and while it isn’t 100% accurate it is a good way to get a starting cap tube length for beginners. I myself just go by what I feel is a correct length for the system I’m building, most of the time I’m pretty damn close and might have to trim a few inches after I tune it a few times.I have plans for reworking this entire article though as it’s a bit dated and it’s terribly organized.

  49. blindfitter {Tuesday July 24, 2007 @ 10:27 pm}

    Hi Chris,
    Great piece of work, confirmed a lot of my queries, but added a couple of questions,

    1: Why on insulation, leave the last 2″ uninsolated? mine if I do that frosts up, do I have a problem?

    2: Floodback, is this a gurgling sound?

    cheers and thanks


  50. Chris Morrell {Tuesday July 24, 2007 @ 10:43 pm}

    On some of my builds I insulate all the way up to the compressor but on average I leave an inch or so uninsulated. Under load usually this portion sits right around ambient as the heat from the compressor creeps up the suction line. I do this as it’s nice to be able to quickly see if you are heavily flooding the evap and suction line. Floodback sounds like popcorn popping within the compressor shell, it’s very distinct. A little liquid return isn’t all that bad, but enough to make it crack and snap will lead to a dead compressor.

  51. Dr. Mario {Tuesday May 12, 2009 @ 6:19 am}

    Hmm. Interesting project - THAT was what I wanted to do. However, I may be doing that someday as soon as I can find the DC brushless compressor to tinker with.

    And, on other notice, on-die memory controller is what Athlon 64 used to smash Intel Pentium 4, at stock level. The speed is entirely limited by RAM’s operating frequency too. Higher LSB (RAM FSB), the faster your AMD64 processor runs. In theory, if it uses 3.2 GHz XDR-DRAM, it would go from 2 GHz to 7.4 GHz chilled well. Try AMD Phenom II too.

  52. Robert {Saturday June 27, 2009 @ 3:01 am}

    Hey Chris, thanks a lot for the article! I’m doing the submersion cooling project (puget custom comptuers and tom’s hardware have videos on it), and so I need to cool the entire tank of oil. I was having problems because it’s supposedly a 430W load (weaksauce PSU xD), and when I use your rule of thumb when picking a capillary tube length, i wind up with -8 feet. Should I make the tube 8 feet in the other direction, or what? >.>
    Hehe, maybe I’m not understanding this right. Thanks


  53. Robert {Saturday June 27, 2009 @ 3:08 am}

    Hey, Robert again- I suppose I should mention that I’m not particularly interested in sub-zero temperatures, just sub-ambient (5-25 oC). Maybe chilled liquid cooling would be better? =\

  54. Thomas {Thursday August 20, 2009 @ 3:09 pm}

    Very Impressive!

    As brazing with shielding gas is very difficult / expensive for most of us, what about bonding with epoxy resin? I heard that it is already used for repairing refrigerant tubings where brazing is not possible.
    What makes me wonder is that resin is not stronger that plastic and so maybe not really 100% gas-tight.

    Does anybody know if the problem with soldering is long term mechanical stability only and not also chemical reactions of the solder (mainly tin?) with copper and refrigerant?
    If so, one could tape a soldered joint of pipes with a bushing of glass-fiber and epoxy to make it as stable as brazed.
    Would i still have to use shielding gas when soldering or is temperature low enough that oxidation does not occur?

  55. Algoulsk {Wednesday November 25, 2009 @ 12:37 pm}

    Great post, but got a q: the formula you posted, concerning the wattage based on the capillary tube, how do you choose the wattage? should you be precise or just all out crazy n go for the max logical wattage? I am building a phase change/ anti freeze water cooling hybrid, with all the items in my pc being cooled, ie psu, cpu, gpu, ram, mobos mosfets, gddr ram, etc. but i don’t know what wattage to choose? help? I don’t like the direct evaporator method so i’m opting for the hybrid idea.

    Any advice anyone?

  56. Teach Yourself Violin {Wednesday December 16, 2009 @ 10:45 am}

    Hah, I think that this may definitely be a little bit of overkill.

Speak Your Mind

  • Comment Policy: No flaming unless directed at the author for good reason. All other flaming will be turned into a comment about ponies and rollerskates.