Exploring Below Ambient Water Cooling

8/31/2012 Thread Update Note: This is a repost of the original thread, this cooling was actually put into use in June of 2011.

As the title says, "Exploring Below Ambient Water Cooling", below is the cooling setup that was in use when this thread was first posted, the pictures are still here simply because it is a viable fully functional alternative cooling solution. This cooling solution has gone through many changes testing different ideas you will need to scroll into the thread, to discover those changes, as they reference what works and what doesn't.

Scrolling close to the end of the thread, to see the present setup using TEC (Peltier), assisted cooling eliminating the ice altogether, taking this cooling solution from high maintenance to literally zero maintenance. Use this information however you decide to apply it for your own needs and goals. This is definitely exploring below ambient cooling, enjoy the pictures and the read, and thanks for checking the thread out! Ryan

Link to TEC/Peltier CPU Water Cooling Thread

I have run traditional water cooling closed loops, but the traditional closed loop did not meet my cooling needs, I was looking for a cooling solution to use below ambient water temperatures, to handle the overclocking load temperatures I was reaching, it's given fact you can get more 24/7 overclock stability, the cooler you can keep the CPU.

Pictures can say a thousand words, and from the very beginning of this project there have been changes and modifications made to improve the performance, solve issues and problems, and being this is a below ambient water cooling solution it needs to start with a Caution.

Caution: Water temperatures down to 0c produce condensation, and if you intend going that low motherboard insulating precautions have to be taken.

You can safely operate approximately 8c below ambient room temperature, more than likely wherever you live without any condensation at all. That 8c below ambient is an extremely conservative number, as where I live it's 13c below ambient that I run. However that is totally dependent on the dew point level where you live, and you will have to test and discover how low you can safely run below ambient, without condensation, for yourself.

More regarding temperature and condensation below.

Also this is not a mobile solution, you will not be carrying this cooling solution to your local lan party, even though you could always tap into your buddies drink cooler. :lol: JKing

So speaking of cooler, even though the pump is the heart, the cooler is the reservoir in my situation holding 10 gallons of distilled water, leaving enough room to add if needed 2 2/3rd filled jugs of frozen distilled water, ICE blocks, to drop the water temperature to where I want it.

I'm using a Coleman cooler widely available even at Walmart, the model I am using and most Coleman end drain cooler models have the perfect size drain outlet to accept a barbed brass fitting, with nothing more than a hose clamp to secure it.

Feed Line Connection Close Up

The water return line had the be cut in so the lid could be closed and sealed, the wire you see in the picture is a temperature probe.

Return Line Close Up w Temp Probe Wire Entering Cooler

The heart of the system is a Swiftech MCP655 variable speed pump.

All the tubing is insulated.

Critter control by AC Pan Tabs, 2 tabs per gallon have worked nicely, no critters growing in the solution, no corrosive metal or plastic damage, these AC Pan Tabs are stocked at Home Depot or can be ordered online. Add the tablets 2 at a time allowing complete dissolving between additions until you've completed adding all you'll be using. Do not add all the tablets at one time! I also run 1 Silver killcoil inside the pump intake tubing, located as far away from the pump as possible.

Pan Tablets Package Close UP

The perfect filters for the Cooler Master drain which is the pump intake line are the Spectre high performance gas filter replacement cartridges pictured below, stocked at your local Autozone or also orderable online.

Filter Package Closeup

They are nylon mesh open end filters that you cap one end and the other end fits snugly into the drain opening.

I have since modified the filtering to allow longer run time between filter maintenance swap out and clean by mating two of the filters together and used JB Weld epoxy to do it.

Since you have to open the cooler to change out the ice blocks, dust in the air gets into the cooler, the filters trap any dust particulates, and keep them from entering the pump and water block, keeping the circulating part of the system clean.

Use whatever water block you like this is an XSPC Rasa and a Danger Den.

Temperature and Condensation:

Humidity levels where you live, will affect the operational water temperatures you can drop down to and safely run with condensation not a problem. You will have to do your own tests to discover what water temperature level, condensation begins to occur. When condensation begins to form, unless you take serious insulation precautions, you need to operate at temperatures above that point.

Directed airflow on the condensation producing elements inside the computer will evaporate the forming condensation allowing you to go below the condensation forming point to a certain extent, but eventually you will discover a lower temperature the fan won't stop the condensation.

Condensation begins to form for me at 8c water temperature that's 15c below ambient room temperature, that's some serious overclocking headroom., I usually operate around the 15c water temperature range, which is 8c below ambient. Over time you will learn what works good for you, whether surfing the net, benchmarking, gaming, etc., and how much ice you'll need to support the activity at the overclock you're running.

CPU Temperature Results:

These comparative tests below were run with my 2500K overclocked to 4500mhz @ 1.325v with one 580GTX at 23c ambient.
The 4 core temperatures were averaged together for one total score.

Air Cooling;
Noctua NH-D14, Idle = 32.75c, Load = 54.75c
Thermalright 120 Extreme 2 fans in push/pull, Idle = 32c, Load = 51.75c

Standard Water Cooling Closed Loop;
XSPC Rasa RS240 kit, Idle = 30.75c, Load = 50c
XSPC Rasa with a Black Ice 240 Radiator, Idle = 30.25c, Load = 48.75c

Ice Block Chilled Water cooling;
Water Temperature at 23c Ambient, Idle = 27.75c, Load = 46.5c
Water Temperature at 20c, No Condensation, Idle = 26c, Load = 43.5c
Water Temperature at 15c, No Condensation, Idle = 21.5c, Load = 37c
Water Temperature at 10c, No Condensation, Idle = 15c, Load = 34c
Water Temperature at 7c, Condensation Direct Fan Controlled, Idle = 13c, Load = 30.25c

Note Regarding CPU Water Blocks Used: I have tested 3 water blocks, Danger Den MC-TDX (Highest Flow Rate), Swiftech Apogee XT rev2, and XSPC Rasa, the Rasa Water Block, is the best performer temperature wise.

Thermal Mass:

I've wanted to address the subject of Thermal Mass because some here don't fully understand how this works in your favor with this cooling solution, some relating to slush box reference simply stuck a radiator in coolers or buckets with crushed or small formed ice. There are many examples of various types of slush boxes across the internet of tried projects, some even used dry ice and water, and of course it is a short lived solution because ice melts and dry ice evaporates. In the case of dry ice and methanol slush boxes sub zero temperatures are produced causing a host of problems to deal with way past condensation. I found it comical when one of the guys radiator actually froze, what water circulation did he have then. :lol:

Regarding my setup and thermal mass, I have 8 gallons of water inside an insulated cooler, I use the frozen jugs of water because they are a dependable constant variable, it takes time to cool down 8 solid gallons of water, it does not happen instantly, and in the reverse, it takes time to warm up 8 gallons of water. This is not a short time solution as far as water temperature is concerned even with 2500K overclocks to 5,000mhz and 5,100mhz, thermal mass allows those high clocks to be 24/7 rock solid stable, the only reason I do not run those clocks 24/7 is the voltage it takes to run them.

Without this cooling I could not run my 2500K at 5ghz for anything more than a CPU-Z validation, but this cooling allows not only reaching but acquiring stability once getting there. The thermal mass of the water ensures longevity of certain temperature levels that you may be needing to accomplish your overclocking goals successfully. Additionally for longevity the system is constantly running at cooler levels all the time, even if I allowed the water temperature to rise to ambient room temperature, it still bests a standard water cooling closed loop, it opens completely new doors of possibility.

If you have the end cooling results staring you in the face, the maintenance is a fair trade for the cooling gains. I know this cooling solution is not for everyone, it takes up space, it is a high maintenance solution, it takes a time investment, it is not portable or potable, you cannot drink the water, but if you do have the room for it and you ever try it for yourself, I'd love to know what you have to say then.
Rad Box Mod

I decided to create a Rad Box to handle the GPUs since I already had almost everything on hand to do it, from past water cooling loop solutions. I used an old Super Micro server case, gutted it and added a 360 radiator top, and a 240 radiator bottom. Both radiators are running shrouds which allows full uninterrupted airflow pulled through the entire radiator cooling fin face. The rear 120mm fan is supplying fresh air into the Rad Box and is speed controlled to the point that it's raw delivery pressurizes the box enough that after the radiators have pulled what they need there is still air exhausting out the front intake, so it is supplying enough air for the radiators to use plus.

Extended the case feet to allow the bottom fan enough exhaust height, bonded the extenders with JB Weld epoxy the black tape is temporary.

Top Radiator View

Bottom Radiator View

View of XSPC Pump/Res and the fan controller.

Inside View.

Outside View.

All lines from Ice Cooler to CPU, and Rad Box to GPUs, are fully insulated.

Rad Box Lines to GPUs.

Ice Cooler lines to CPU.

Full View inside computer.

Rad Box Finished View.

Top View

Fans Close Up

Angle View

Just for Giggles > Ry's Rad Box ;)

Rad Box GPU cooling testing:

For the record; The stock temperatures of the air cooled 580GTX cards were Idle = 34c, Load = 80c.

The 580s cooled by the ice cooler with a water temperature of 19c, Idle = 23c, Load = 35c.

The 580s now cooled by the Rad Box, Idle = 27c, Load = 40c.

The Rad Box for cooling the GPUs is definitely successful, it totally removes the GPU added heat from the Ice Cooling, which has allowed dropping the Ice Cooler pump speed to level 2, and now 1 jug of ice in the cooler does what it took 2 to do without the Rad Box.


Thanks for the updates ryan. I fell in love with this setup when I first saw pics of it. The ability to add in jugs of ice is what I love most about it. Great idea and I'm pleased that it seems to be working so far.

How much heat do you think you put out? How many jugs do you need to keep in the freezer to have two in there at all times? Finally, what temps do you see with ice and without?


Mar 26, 2012
you do realize that your begging me to build that freezer unit, to show how it works right. but back to this setup are you rotating the ice gallons out every 4 hours or so, and how big is the freezer the all of your gallons are freezing in, ok maybe that is too much. but how long can you sustain your clocks before you have to bring it back down to idol. I only ask because this seems like a plausible transition phase between now, and latter after I build it. man, I like to game, and at high settings you get to feel the game the way the programmers intended it to be played. so really the reason I am so interested is for quite selfish reasons, but I hope you can understand that I do really like this idea.
The Pics Below are a Lesson Learned the Hard Way!

This is part of my learning process, I'm sharing this with you all, however the pictures below are not presently in use at all regarding the project, they are a part of the learning process journey. The Rad Box is a much more efficient cooling means for the GPUs, and is fully operational and in use. This project has been a long road of discovery and the things I have learned, I am sharing here in this thread, so if any of you decide to use this cooling solution, you'll have this information for reference.

Just cooling the CPU I was using 6 jugs total, that clarified is at the most 2 jugs in the cooler, 2 completely frozen in the freezer (The next to be swapped out), and 2 being frozen in the freezer.

Typical daily use

If I was just surfing the net and such, 1 jug in the morning, 1 jug in the evening, would keep me around 15c.

After adding the GPUs to the cooling loop, the added heat to run the same 15c temperature gaming, required 2 jugs for each change out, and forced me to change the 6 jug rotation to 10 jug, increasing to 2 jugs in the morning, 2 jugs around noon, and 2 jugs in the evening.

Gaming is a serious heat increase with the GPUs added to the mix, with an overclocked Sandy Bridge K series CPU.

Keep in mind these jugs are only 2/3rds filled because expansion of the water freezing will split the jugs open if completely filled, additionally freezing and refreezing stresses the jugs and they'll only freeze and refreeze approximately 10 times and begin to leak, that's why I use distilled water in the jugs so a leak will not contaminate the water.

When I added the 2 GPUs to the cooling to meet my cooling needs I did another modification pictured below and split the return line to the cooler and added 2 radiators in the loop I already had on hand from previous water cooling loops as a side cooling option.

Using the radiator loop itself for the total cooling took me back to closed loop water cooling performance which I was never satisfied with in the first place. It also brought new problems to the table, the water temperature in the cooler rose to 26c, 3c above ambient, and was the perfect environment for mildew to begin forming under the cooler lid.

I quickly aborted that idea, cleaned off the lid, closed the radiator line off and went back 100% to the ice cooling.

There were no problems with the water itself in the cooler, just the air space above the water inside the cooler, was the perfect temperature range to support mildew growth inside the cooler, with the water temperatures below ambient, that problem was non existent in the cooler, it started from the increased water temperature.

So I needed a cooling solution for the GPUs themselves, putting them on a traditional closed loop with the smallest amount of air possible in the reservoir, seemed the most viable solution, and when toying with the idea of how to best accomplish that, I had not considered a Rad Box until seeing Moto's, and decided that was the best route to take, Thanks Moto!

Using a Rad Box for the GPUs returned my Ice Cooling of the CPU back to the way it was in the beginning, which has brought my ice use back to the original 6 jugs being cycled from the freezer to the cooler.

Here's a major actual fact learned from this!

Using radiators in a chilled water setup is a bad idea, Period!, (The radiators have a reverse effect and heat the chilled water attempting to bring it back up to ambient!)

Let me say that again!

Using radiators in a chilled water setup is a bad idea, Period!, (The radiators have a reverse effect and heat the chilled water attempting to bring it back up to ambient!)

One more time!

Using radiators in a chilled water setup is a bad idea, Period!, (The radiators have a reverse effect and heat the chilled water attempting to bring it back up to ambient!)

Please take my reply to 4745454b into consideration as some of your questions were answered while typing it up.

I actually have 2 freezers in my home and the jugs are spread between the 2, moving the GPUs to their own closed loop as said above, will remove their added heat from the cooler.

This cooling solution has allowed a 5.1ghz CPU overclock of my 2500K with rock solid stability, and even though I could run it daily 24/7 at those clocks I run it at 4500mhz, to run lower voltages over time.

I can still do the same with the GPUs added it just takes much more Ice for gaming.


Contributing Writer
I don't think anyone on this forum really can say that anything or any idea is crazy.

We're the guys that run water to cool their computers...that might put us in a room of our own, but we're all on the same boat, in the same padded room and all sharing the same straight jacket.

There is never enough crazy when it comes to watercooling.

Edit: Have you considered a circulation pump for your cooler? I wonder if this would increase cooling performance much like a convection oven with circulating air improves consistent baking?


Mar 26, 2012
ok that is cool, I don't have two freezers, but I do have one stand up freezer, and it can hold about 30 gallons. do you think that will get me to about six hours of game time or would that be three.


Contributing Writer
He's using a large cooler, not a freezer or refrigerator to cool. You wouldn't necessarily use a normal refrigerator or freezer as-is, unless you modified it. I think we had this discussion on another thread once, so while I'm not trying to say it's absolutely not plausible, it's really not likely with a fridge or freezer at stock config. He's freezing the ice jugs in his freezer and then placing them in the cooler to drop the water temp down for higher overclocks and lower temp stability.

Good point, overclockers are a minority in the computing realm as a whole, and water coolers are a minority of the minority.

30 gallons?

You need to read this thread again.

With 8 gallons of water in the cooler I can only add 2 jugs of ice at one time?

If I drop the water level to 7 gallons I can add 3 gallon jugs of ice, but that will drop me down to condensation producing levels and I don't want or need to go that low.

Any cooling you can attain below ambient room temperature even just 3c below ambient changes overclocking headroom drastically.

20c is 3c below my ambient room temperature, I use 20c as my change out point and swap to 2 freshly frozen jugs, 30minutes later I am at 15c and it takes about 4 hours to get back to 20c.

I can game all day long if I want even with the GPUs added to the mix swapping out 6 gallon jugs 2 at the time, that leaves 4 fully frozen in reserve still in the freezer.

FYI, I don't game for 12hours at the time, I don't know your gaming habits, but maybe this clears it up some.



Mar 26, 2012
I realize that, and I realize that the hot water coming off the system is going to melt the ice pretty quickly. that's why I asked how many hours will 30 gallons get me at 10 gallons per fill or at each switch as temps start to rise.

ok I thank I get it now, you're saying 10 gallons total. so that's 5 gallons a day, or something to that effect.


Contributing Writer
In all honesty, what you are asking is dependent upon far too many variables in order to correctly predict. Essentially this kind of project is the same as running a very large reservoir and adding ice to cool water below ambient temperatures. The ice is cooling the heat output as well as lowering overall temps lower than ambient since radiators are not in-use.

Radiators are going to be in use, this is the beginning of an entirely new update, I've built a Rad Box specifically to cool the GPUs.

Missed that, running the return line to the rear of the cooler serves the same purpose.

It's a perfect GPU cooling solution, and a lot neater.


Contributing Writer
I stopped to think about the circ comment...the flow itself would be enough to circulate. Not enough caffeine yet this morning.

What issues came up with the GPUs and rads in the loop? You mentioned there was an issue.


Heated the water too much, caused growth. I'm not sure I would have gave up. Perhaps some more rads before it hit the water? Or change the jugs more often? Or just flat use growth killer...


Contributing Writer
I personally would use heat exchangers with a closed loop, and use the cooler as a water/slush box with the rads submerged inside. The WC loop would remain closed, but keep either a coolant, or growth controlled water in the cooler and use the ice jugs in the same manner.

The cooler internal growth has to be from condensation collecting and then harboring crawlies, otherwise, I don't think anything would grow in the coolant he has cooked up using those microbial inhibitor tablets.

You've stated many times you personally would use heat exchangers with a closed loop, even when I first posted this cooling solution over 11 months ago.

Have you ever actually used your own suggestion?



Contributing Writer
I have once, several years ago with a bucket, ice and a heater core when I decided to see how well it would work on my P4 system when it was headed to it's afterlife.

Just trying to offer ideas- I know I've mentioned this before, but there are several others who might not have read through your setup/config that may consider it if they were considering a build such as this.


I have not given up, and adding more radiators would not have solved the problem, what I needed was simply a dual loop solution, because the GPUs actually do not need the ice cooling in the first place, they are perfectly fine being cooled by a radiator closed loop.

The growth I was encountering was from condensation accumulating underneath the lid, and the warm moist air environment, of the large air space above the water in the cooler, when the water temperature reached 26c that's 78f, a warm enough environment for the mildew growth to begin.

I wasn't cooling with ice when it was completely being cooled with the radiators, although I did try that approach, unfortunately the radiators themselves had a reverse heating effect running chilled water through them, they actually heated the chilled water faster.

Even with the water temperature reaching a standard loop temperature for my setup, my GPU load was still 40c below the 580GTX stock air cooling solution, so that's the same cooling I'm expecting using my Rad Box to cool the GPUs.

Removing them from the CPUs loop, once again the CPU will be back to the only thing being ice water cooled, returning me to my original goals of high multiplier overclocking versatility with my 2500K.


Contributing Writer
I wasn't cooling with ice when it was completely cooled with the radiators, the radiators had a reverse effect running chilled water through them, they actually heated the chilled water faster.
I was wondering if this was actually what was happening... the ambient air would likely have a much larger impact due to the total CFM of air and ambient heat in watts being exposed to the rads vs. what the cards would be dumping into the loop and needing to be cooled.


Contributing Writer
Considering it was done years ago and mainly just to see what would happen, it wouldn't scale well to today's hardware standards without some changes in heat exchanger surface area/volume and updated heat loads. The main reason I've mentioned it was due to the similarities of the project.

I was seeing condensation on the CPU block, which required me to maintain high airflow around the block along with some paper towels to help absorb the water. I don't recall there being an excessive amount to the point of drops dripping, but it was noticeable. I don't think I was using any sort of temp monitoring...maybe ASUS Probe, which in itself was far from accurate. I don't recall if there were a lot of viable temp monitoring tools that I knew of at that time, to be honest.

The main reason that I don't use this as a cooling solution now (scaled up, obviously) is simply from a standpoint of practicality and mobility. I occasionally go to LANs (although much less over the past few years) or even to do some gaming at a friend's house instead of all being remote and online. I also do not want to maintain an ice supply in order to cool the water in the slushbox. I simply had a 5 gallon bucket and a bag of ice from the grocery store for my testing/experimentation I did long ago. I know that I would obviously need a different approach with a long-term design, but I have no desire to constantly maintain ice for this purpose. True, there are other alternatives for 24/7 operation in addition to this design, but I'd rather have the 24/7 without the downside to the 'when I want the ice' part of the time.

I've never questioned that your cooling solution didn't work; my main point of offering ideas or suggestions was to further the options for other forum members reading about your build. The original thread was either deleted or is lost in the forum somewhere, so while I know I have mentioned the heat exchangers a few times, not everyone has read the old thread to have seen them.

The DIY factor and results speak for themselves as to how well your setup is working. I think that is overwhelmingly established.
Thread starter Similar threads Forum Replies Date
LordTimothyDexter Overclocking 1