TEC/Peltier CPU Chilled Water Cooling

Status
Not open for further replies.
This thread is sharing information with you regarding Peltier/TEC CPU cooling, it is to bring those interested, or curious regarding this type of cooling, up to date on it’s cooling possibilities, and progress, and the setup pictured below is currently in use and fully operational.

Caution: Peltier cooling will allow you to go below ambient, depending on what your desires and goals are and how many peltiers you run, you can even go below 0c, it is possible using anti-freeze coolant. My initial goal was to be able to run below ambient just above the condensation forming point so motherboard insulation does not come into play. You can safely enjoy about a 13c below ambient without any condensation worries at all, but you take full responsibility with what you do with this information.

100_2492.png


100_2493.png


100_2494.png


100_2504.png








100_2498.png


If you’re interested in following the full journey of how this all came to be, it is in the Exploring Below Ambient Water Cooling thread.

What does this type water cooling allow:

I’m sure the first question would be why would anyone be interested in this type of cooling in the first place, and if you’re 100% satisfied with your present water cooling setups load temperatures, you won’t be. The main why is, it can run load temperatures, lower than the mass majority of your ambient reliant CPU big air coolers, CLC Coolers, Standard and Custom water cooling loop idle temperatures.

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, and was run on WinXP 32bit.

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

Tests below run with a 2700K with hyper threading disabled, which comparatively temperature wise is equal to a 2500K, all the other test parameters are the same and conducted on the same motherboard.

TEC/Peltier Water Cooling
Water Temp. is 12c which is also, 12c below ambient room temperature, (zero condensation), Idle = 14c, Load = 31.25c
Water Temp. is 9c which is, 15c below ambient room temperature, (zero condensation), Idle = 12c, Load = 28.5c

2700K @ 4500mhz Win7 64bit, Intel Burn Test

2700K @ 5300mhz Win7 64bit, Intel Burn Test

3770K @ 5000mhz Win7 64bit, Intel Burn Test

Peltier information:

In my own words, The peltier is a simple device, you run electricity through it and the reaction between the two dissimilar metals results in one side getting hot and the other side getting cold, if you do not cool the hot side it gets too hot, overheats, and burns up, simple as that. The Peltiers cold side thermal output can be harnessed and used. In this case the cold side is being used to chill water flowing through a modified water block and storing that chilled water in an insulated reservoir, which is then used to cool a CPU.

The peltier itself is essentially a miniature freezer/heater with no moving parts to break down, the key is discovering how to use what it can do, to your advantage. presently the hot side exhaust is being used to heat my office and in the winter it does a very good job of it, in the summer I have to run my office AC anyway so it counters the added room heat from the TEC assembly.

More detailed peltier information, credit to xtremesystems.org.

When the CPU (overclocking), cooling community first put the peltier into operation it was directly mounted on the CPU usually water cooling the hot side, which brought unpleasant side effects requiring motherboard insulation to keep from shorting out components from the ice forming around and on the backside of the motherboard socket.

Ice occurs when the surface temperature is cold enough to freeze the moisture in the air, the more humidity in the air the worse it would be.

Comparatively those earlier CPU peltiers were only a small percentage of the size of the peltier I am using in this project, a Potted Peltier, measuring 50mm x 50mm x 3.10mm, wattage = 226w ~ 245w, amperage = 26a max, voltage = 12vdc ~ 15.4vdc, with a maximum operating temperature of 125c.

Note: Potted means insulated, some peltiers are not insulated at all and really are not suitable for this type of cooling, you would want to acquire a potted peltier.

Many at Overclock.net use direct to CPU mounting methods today and if you're interested in that type TEC application you can research there.

This thread is totally about using the Peltier/TEC assembly to replace the ambient limited radiator cooling, with a chilled water cooling solution allowing going below ambient room temperature.

It would have been nice to have received some kind of instructions with the peltier, at least which side produced hot and which side cold, I had to use a 1.5v battery to determine that, I guess they figure if you buy something like this in the first place, you should be smart enough to figure out how to use it, anyway still, instructions would have been nice.

When energized this peltier and power supply pulls 277w by itself, so whatever total load you are running this is an added load the entire time it's energized. Depending on the voltage the load temperature of the peltier, (and load temperature is when voltage is applied period!), and it varies from spec'd model to spec'd model.

Logically it would seem the best solution would be to keep the hot side as cold as possible, but that's not the case, you have to allow the hot side to get hot for the cold side to get cold, but since we're only talking a approximately 1/8th inch thick plating, that's a very close distance to play the hot/cold balancing act.

Some of what needs to be shared is the peltiers what I call weirdness, I originally thought the cooler I kept the hot side the cooler the cold side would be, however the hot side has to get hot for the cold side to get cold, it is a perfect example of, "For every action there is an equal and opposite reaction.".

There was quite a bit of testing and fan changing to get the peltier to run at it's optimum cooling side capability, at one point I actually had the hot side too cool and it was not performing as expected, it's been a learning experience that's for sure.

Simple discovery the hot side can override the cold sides effectiveness but news flash the opposite can also happen, the cold side getting too cold, can override the hot sides effectiveness,

So you're after a balancing act of hot and cold to get what you're after, the true key to this things performance is give it the heat range it operates best within or sweetspot, and it will give you the results you're looking for. It needs to be in the mid range of it's capabilities, so it's hot enough to deliver the cooling performance, but far enough away from it's limitations to keep from burning it out.
 
The TEC Assembly consists of the Peltier module sandwiched between the heat pipe cooler with a cold plate to reach the full coverage of the Peltiers hot surface and the water block that picks up the cold from the peltiers cold side.

CPUpeltier3.png


Close Up

GPUpeltier4.png


Power Supply Information:

A single 12v rail is the best PC available 12v output power supply to use to power a peltiers constant load, make sure it has the raw amperage handling capability, well on the plus side. With a PC capable P/S 12v, you will need to get power from 14g main wires, like from the 8pin or 4pin motherboard connector, those are usually 14g.

Danger: Do not use the 18g 4 pin molex connectors as they cannot handle the load, they will burn up and take the power supply out, and more than likely anything else hooked to it. Additionally do not compromise this power supply selection, to be on the safe side you need amperage capability higher than the peltiers load.

When it comes to powering a peltier you have to get your power from an output area of the power supply that can handle the amperage load. I used 2 yellow positive 14g wires from the 8pin motherboard connection line soldered together, as the positive primary, and 2 black 14g negative wires soldered together as the negative secondary to power mine.

You do not have to use a power supply readily PC available you can find online specific voltage/amperage rated models that will be specifically what your peltier requires, the challenge with those is getting them to power when the PC powers up or down.

100_2471.png


100_2497.png


The power supply in the TECBOX is jumpered to the power supply in the main computer, when the main computer is Started Up or Shut Down the TECBOX power supply does the same.

Cooling the Hot Side:

Cooling the hot side of the TEC is a challenge in itself, seeing as how it's best performance comes from allowing the hot side to get hot, however the maximum heat range for my TEC is 125c which is 257f, way past the temperature to boil water. I originally thought water cooling the hot side of the peltier was a good idea since in it’s earlier days it was water cooled but much smaller peltiers were being used, nowhere near these capabilities.

In my attempt to water cool the hot side, the radiator got so hot I couldn't put my hand on it, the XSPC water pump got so hot I burnt my finger on the metal output spout. At that point the peltiers hot side temperature reading on the outside was 160f, I estimate about 180f in the center of the peltier, cooling that much heat is serious and in itself takes a serious dependable and capable solution.

With water cooling the hot side of this higher amperage peltier off the possibility table I turned to heat pipe air coolers, the first was a shelved Tuniq Tower (the original model), with only 3 heat pipes, it just could not overcome the heat the CPU was adding to the water.

I had a Thermalright Ultra 120 Extreme on the shelf and used it to cool the hot side of the peltier, it has 6 heat pipes and is very effective at handling the heat produced by the peltier, also an old Xigmatek S1283 HDT cooler works very well.

Experimented with multiple combinations of fans in push/pull configuration and settled on 110cfm Sanyo Denki San Ace 120mm fans on a Sunbeam fan controller, they give the best of both worlds, they're set low speed presently you can barely hear them but when needed they can crank.

A good heat pipe air cooler can and will cool the hot side, however my choice reasoning was to choose coolers that could still handle the heat load if the cooling fans completely failed. A heat pipe cooler that is on the cooling performance level of the TRUE (Thermalright Ultra Extreme), or the Noctua NH-D14, is sufficient remember the TEC will burn up if it is allowed to get too HOT.

CPUpeltier1.png


GPUpeltier1.png


Water Block for Cold Side:

Todays water blocks like the RASA are very flow restrictive, the dispersion nozzle is right down on the base plate pins, it's designed to drop it's cooled water directly on the hot base plate forcing it through the dispersing pins.
A water block is designed completely opposite of what I needed it to do, it is designed to pinpoint cooling to the CPU die location, I needed it to gather cool from the entire copper base contact area, as it was exactly the same size as the peltier.

The cold side of the peltier can produce - 0c temperatures, that's not a time for a block that seems to love getting fouled easily and dropping the flow rate, to allow the water to stay in the block too long without flowing, or it will freeze the water block and split it wide open.

I modified the RASA water inlet injection protrusion inside the top of the block and shaved an 1/8th of an inch off the depth, and also cut diversion channels in it to quickly disperse the water to the entire base plate, transferring the cold from the baseplate which contacts the cold side of the peltier to the water, and forcing it out the outlet.

This water block modification made it very effective in picking up the cold from the peltier transferring the cold to the base plate.

The Thermalright (TRUE), is running a Swiftech Apogee XT rev2, and the Xigmatek running an XSPC Rasa, both water blocks have been flow rate modified.

Note to anyone considering duplicating my setup:

The Swiftech Apogee XTL is the best block for flow modifying to use with a 50mm x 50mm size peltier, I'm recommending it over the XSPC Rasa, the Rasa will still work, but not as well as the Apogee XTL.

Due to the XTLs larger copper base plate the entire 50mm x 50mm peltier fits in solid contact fully covering the contacting face, whereas the Rasa base plate is exactly the same size as the peltier, requiring filling the screw indents with thermal compound.

The Apogee XTL is exactly the same design of the original Apogee XT, but the top is machined from black delrin, which is much easier to cut and modify, and is also cheaper, the copper base is wider allowing further coverage and cold pickup of a 50mm x 50mm peltier.

I must start this next comment with apologies as my camera is malfunctioning so I have no pictures to show of the modifications I did to the XTL to increase its flow, but it works very well.

I additionally had a HDD failure which cost me every picture I had taken that had not been backed up, we tend to take things for granted since things seem to be running so smoothly, but that failure beast is still out there, make sure you backup regularly what you do not want to loose!

I will be taking new pictures of the setup as it sits now and replacing the pictures in the beginning of this thread once I either discover what's wrong with the camera or use another to take them.

PerformancePCs is one of the only places you can still acquire the Thermalright TRUE without the fans, which is great if you already have the fans you need.

They also have an excellent price on the XSPC Rasa, if that is the block you prefer.

Picture below is the Swiftech water block assembly.

100_2502.png


Picture below is the Rasa water block assembly.

100_2500.png


100_2505.png
 
I modded my Rasa Water Block much more than I had remembered!

RasaMod2_zps82b2039a.jpg


RasaMod1_zps75f66856.jpg


I remember starting with the wagon wheel design, but had the block begin to freeze up, not to the point of splitting it open, but to the point of freezing up enough to choke off the water flow, so I obviously modded it again.

This is the end result and what I've been running.

My apologies for the pictures!

My digital camera is giving me a fit presently, I took about 20 pictures just to get these 2 and I'm not sure why it's happening?

This camera has been fantastic up to this point I fear I may have to replace it.

RasaMod3_zps6c7c22ff.jpg


Major changes from the original the first step protrusion was completely removed.

RasaTop_zpse1f23e6b.png


Insulated Reservoir:

100_2481.png


100_2499.png


100_2495.png


After many trial and error experiments I settled on an insulated reservoir I made from 4" PVC schedule 40 pipe, insulated with aluminum faced foam duct seal, as the best solution to arrive at the proper balance of water mass vs peltier cooling capabilities. From the beginnings of this testing (Referenced in the Below Ambient Thread), using a 54qt 13.5 gallon down to the present ½ gallon or 2 litre capacity. Using a lower volume of water reducing the thermal mass, allows faster cooling transfer to the water.

The insulated reservoir becomes a cold water storage tank which becomes the buffer the TEC cooling needs, allowing the peltier/TEC assembly to counter, and lower, the heat generated by the overclocked CPU. Additionally all the tubing runs are insulated to retain as little cold loss to the ambient room temperature and heated operational elements inside the machine.

100_2508.png


100_2496.png


100_2486.png


Performance Variables :

Water Flow Rate :


Water flow rate affects the cold conduction of the peltier to the water flowing through the water block, too fast and the water does not get cold fast enough, too slow and the block can freeze, presently set on level 2 of 5 settings using a Swiftech MCP655 pump. Flow rate is also a consideration when balancing out the system, regarding taking into consideration the CPUs water block flow restriction. The perfect balance is having the peltier water block able to completely counter the heat the CPU is adding to the water.)

Peltier clamping pressure:

I started with about approximately 25psi ~ 35psi estimated which is basically a standard heat sink clamping pressure, then increased to approximately 50psi+ then after discovering on the net even higher pressure was needed, increased to approximately 100psi.

I'm saying approximately 100psi because I have no actual way to gauge the actual pressure it is wrench tight though, way past any spring clamping.

Addressing Condensation :

Where you live has everything to do with at what water temperature condensation begins to form on the CPU water block, regarding relative humidity, and dew point. Where I live condensation begins forming as a frosty look at 8c water temperature which is 15c below ambient and 16c ~ 18c below the best water cooling radiator loops and as long as I stay above an 8c water temperature condensation is a non issue.

I have an AC unit in my office, that tends to keep humidity levels down quite a bit and allows lower non-condensation freedom. Additionally you can use positioned cooling fans to evaporate the condensation if you need to go lower but by the time you drop down to about the 3c range of water temperature the condensation builds too fast even for fan drying. The fans can actually blow drops of water from the block, and of course that is bad!

Juggling the Hot and Cold:

The cooler/reservoir also allows running the TEC at a cooler hot side operating temperature, and that is controlled by the fan speed on the heat pipe cooler, I know that sounds weird, none the less it's 100% true. Getting what you want from the TEC is a balancing act, you have to let the hot side get hot, for the cold side to get cold.

That is a testament as to how good the Thermalright TRUE actually is, that it can make such a cooling difference and control the hot side temperature with no voltage adjustment to the TEC just controlling the fan speed air flow.

Radiators and Peltiers in the Same Loop:

Many had suggested that would be a good idea which turned out to be a bad idea, the bottom line is the radiator raises the temperature of the chilled water, why go to the trouble and expense to chill the water then allow a radiator to undo it. It seemed like a good idea at first, the radiator could add some stabilization to the loop when actually, it did quite the opposite.

So been there done that, with that idea, and if you're thinking it at this point, forget it.

There's no need to go into any in depth explanation of the whys and wherefores, it is just counter productive to a chilled water cooling period!
 
Update:

I replaced my GPU RadBox setup with a Watercool MO-RA3 and mounted it to the back side panel of the TecBox, the 2 580GTX with full coverage Heat Killer water blocks are now being cooled with the MO-RA3.

I did not take as you go installation pictures because that would have seriously slowed the installation as once I had this massive radiator in hand, I was like a kid with a toy on Christmas, I wanted to get it installed!

The MO-RA3 is a massive radiator, it is 15" wide, 16 1/2" high, and 2 1/2" thick, copper tube construction making 4 complete tubing passes through the fin field.

To save me some typing here's a quote from the website.

MO-RA3

The MO-RA3 represents an uncompromising improvement of the successful MO-RA2. Its cooling performance has been increased, even above that of the powerful predecessor. More than 28 meters of tubing and 2.5 square meters of fins speak for themselves. The cooling fluid now flows in a four-time parallel layout through 72 tubes. The fins of the MO-RA3 have been optimized specifically for low-rpm fans. Their enormous surface and effective geometry guarantee excellent cooling performance. The highly parallel layout and the new flow-optimized connection block ensure a low flow-resistance, despite the large size of the heat-exchanger. The connection block improves flexibility as well by providing G1/4“ threads in three directions. Integrated M3 threads allow mounting of up to nine fans per side (not on the MO-RA3 Core LC). Additional threads are integrated for accessories. The MO-RA3 is shipped ready to use. This includes G1/4“ sealing plugs for unused threads, screws for fans, and spacers for mounting it onto a case. The modular accessories of the Mo-RA3 allow for easy customization. Select from various grills, fan controllers, feet, and external mounting brackets. The MO-RA3 is produced in Europe. Development and final assembly happen in Germany, like the manufacturing of most of its components.

I purchased the 4 x 180mm, allowing me to run 180mm cooling fans, they are much quieter than the 5, 120mm, 2,000rpm fans that were being used on the RadBox, the MO-RA3 has out performed the RadBox by 5c.

My graphics load does not go above 40c, the RadBox allowed it to get to 45c, and 40c is half of the stock air coolers temperature of 80c, thanks to the 180mm fans the loudest thing in my office now is the air conditioner.

Here are some pictures of the finished installation.

FrontFull_zps2374cefb.png


FrontRad_zpsa94c212c.png


FrontAngle_zps41e15e50.png


SideView-2_zpsfde47f1b.png


Reservoir2_zps02673690.png


RadInput_zps23386bf6.png


ResBottom_zps657d65c4.png


PumpandRearTubing_zps7ba146bf.png


The MO-RA3 comes with 4 case mounting 3/4" standoffs, but I wanted more clearance than that, so I opted for the mounting brackets which are 1 1/4" off the case for more airflow.

The Reservoir is a Bitspower Water Tank Z-Multi 400, also 4 Alphacool Angled Adapters double 45s, revolvable.

The pump is a Swiftech MCP655 variable speed.

Additionally I am using 4 180mm Phobya 20mm clear plexi fan shrouds, 4 180mm x 32mm Silverstone FM181
High Performance Fans, and 4 Silverstone 180mm wire fan grills.

 
Fabricated Reservoir

Started with a 4" piece of PVC Drain pipe cut to 13" to allow for a pre-calculated internal volume of just over 1/2 gallon or 2 litres.
13x4PVCPipe.png


I needed a mounting bracket so I used a PVC commode 4" test flange with a knockout, I chose this because once the knockout was removed it allowed the flange to go further down the 4" pipe, allowing it to stick out the top as I needed.

ComodeTestFlange.png


Used a 4" PVC Cap to seal off the bottom of the tube.

4Cap.png


I used my mitre box to cut a flat on the side of the cap to make drilling, tapping, and mounting the the barbed fitting much easier.

Flatcutintomountdrain.png


Used PVC cement to assemble the reservoir.

FullViewUpsideDown.png


Inside view.

Insideview.png


Installed barbed outlet fitting view.

BarbedFittingInstalled.png


View of holes drilled for the 5/8 OD Return Line and the Temperature Sensing Probe.

Flangemounted.png


Water Testing for Leaks

Watertest.png


Fully assembled with 4" test cap in the open end which becomes the top when mounted.

TestCapisTop.png
 
Swiftech Apogee XTL Water Block Flow Modifications. ( After the fact)

The block is shown already modified as previously explained all my previous pictures were lost when my HDD failed, I'll be showing the tools used and the bits for the tools used and where they were used and what for.

Caution: Wear eye protection especially when running the Dremel Tool.



The tools used were a Dremel Tool and a Screw Gun Drill



Dremel bits used on the left



Step bit used on the right





The Step Bit allowed boring from the bottom side to widen the mouth without hitting and damaging the G14 Threads of the water block, do this part of the process very slow, with extreme care!





A Dremel Steel Bur cutter was used for the deep path cutting, and initial cutting out and forming of the pathways.





The larger Dremel diamond coated ball for smoothing





The dremel diamond ball taper was also used for smoothing





The dremel diamond cone was used to bevel the outer edge inside perimeter





The Dremel small buffing wheel was used to polish, used medium speed with a light touch.







Finished results



Ready for reassembly





 
The highest recorded clock-speed was 8.429Ghz on an FX-8150, the guys who did it were using Liquid Helium and estimated the CPU wasn't far off 0 Kelvin. I think to achieve 10Ghz there would have to be a change in CPU technology rather than throwing more cooling at the problem.
 
Ryan (sorry for jumping threads didnt feel like cluttering the other one up)

It looks like your running 3 tec's. is that correct?


I have a old k56 lian li case i was looking at using with this setup.
2 243.5w pelteirs cooled by xigmatek Gaia heatsinks and for blocks using ek supreme ltx's


Thoughts are cuting two square holes in the top of the case and mounting the heatsinks there with the WB's on the inside of the case. the loop will be cooling a 2500k and gtx 670 ftw card.
 
I have 3 peltiers but I'm only running 2 in the TecBox, one runs all the time the computer is powered up, the other is controlled by a switch when needed, if I'm surfing the net or posting at Toms one peltier running is sufficient.

When gaming I turn on the 2nd peltier assembly or TEC and that maintains a constant 10c water temperature during gaming.

The TecBox is only cooling the CPU.
 


It will only take 2 for any CPU, one running when the computer starts and one switched on or off as needed, so why would you need 4 are you intending cooling the GPU as well?

If that is your intentions running it all on a single loop is not a good idea you would need a CPU loop and a GPU loop independent of each other, especially if overclocking your CPU is in the plans.

Unless you'll just be running 1 GPU and 1 CPU, my tests have been with 2 GPUs.

Two peltiers can handle either 1 overclocked CPU or 2 GPUs with full coverage water blocks, but not all in the same loop.

You're better off running multiple GPUs on a stand alone Radiator loop as that will be more than adequate cooling for great GPU overclocks at half their original load temperatures.

I suggest doing some experimentation before jumping into the deep end of the pool, balancing the hot and cold of the peltier to get useable results is trial and error the same as overclocking is.

I would not recommend any heat pipe cooler to cool the hot side any less capable than the Cooler Master Hyper 212+, and that would be the absolute minimum, you need a good heat pipe cooler to cool the hot side.

So if you still went with 4 you're actually talking 4 heat pipe coolers, 4 cold plates, 4 peltiers, 4 water blocks, 2 water pumps, and one seriously powerful single 12v rail power supply, (dedicated to run the peltiers), which comes to a chunk of money out the door.

Remember this is not an out of the box solution, it is going to take a lot of work on your part to make this happen.

And you have to have an insulated reservoir for each loop to store the cold water, or I don't care how many peltiers you run it won't work!

Others have tried it without the insulated reservoir and it did not work, and they quit and gave up completely it is easy to get discouraged with this type of cooling you have to have tenacity to persevere to reach a successful cooling solution.

I've covered a lot in this thread of what I have learned using peltier cooling, but the Exploring Below Ambient Thread covers some of the things and ideas that did not work.
 


The rig im going to build/test on is a 2500k and a gtx 670. Thinking of res-pump-tec-tec-cpu-tec-tec-gpu-res or two loops like you said. Its going to make the case heavy but i have two ocz 1000w 83 amp single rail powersupplys i could use or a silverstone 1200w 104a powersupply. the silverstone will handle 3 tec's but i think it would be pushing it for 4. Its only a few bucks more for the Cooler Master Hyper 212+ over the xigmatek Gaia so might as well do that upgrade.

This will be a slow project and will take some time to collect all the parts for this and do all the testing. So right now this is number 3 on the list of things to do, gots a hole list of parts to put on the car and am in the middle of fabbing up a gauge cluster as well.
 


I would do the proposed CPU setup first and go ahead with any overclocking you intend regarding the CPU to see what leeway you have, the overclock you decide to run 24/7 will determine what cooling level temperature wise of the water you'll need to run at.

What are your hot side of the peltier cooling intentions?

 


Thinking Cooler Master Hyper 212+ with push/pull fans with fan controllers
 
Hi all!

It's been a while since I posted any new info so I figured I'd post my latest gaming results regarding temperatures and such.

Been playing Crysis 3 one notch below the maximum settings with the 580GTXs in SLI, the GPUs are radiator cooled and at a 24c ambient room temperature the highest recorded temperature was 43c, vs stock air at 80c.

CPU wise, gaming is the only time that I run the 2nd peltier, and the 2700Ks load temperature overclocked to 4500mhz was 31c, with a 13c water temperature, which is 11c below the ambient.

This setup is running great GPU/RadBox and CPU/TecBox cooled, the GPUs literally run half their original load, and the CPU even on one peltier never reaches ambient room temperature.

I'm still contemplating future performance improvements along with shrinking the entire cooling solution down, but right now it's working so well I hate to even consider messing with it.

Hope you all are doing well! Ry
 


The thread hasn't stopped, I've just been running the setup to time test everything.

It's nice to see someone else venturing into using this type cooling! :)



Ideal?

Could you be more specific?

If you don't mind share as much detail as you can to help others that may be considering this cooling solution, some pictures of your setup would be nice as well.

Could you also list your peltier specs?

The temperatures you're listing are they all from CPU readings?

What temperatures is your actual water temperature ranging?



 


Addressed that issue above under: Power Supply Information

Danger: Do not use the 18g 4 pin molex connectors as they cannot handle the load, they will burn up and take the power supply out, and more than likely anything else hooked to it. Additionally do not compromise this power supply selection, to be on the safe side you need amperage capability higher than the peltiers load.

You don't put changing that problem till later, you need to resolve that immediately, the wires will actually get so hot it will melt the insulation from them and short out your entire power supply, and could actually catch on fire, plus a fan cable may be 20g or 22g wire not even 18g, even more dangerous!








http://www.ncix.ca/products/?sku=64389


???How exactly do you have peltiers hooked to that???

 
Interesting concept but far from the true cooling capabilities you could be getting from the peltiers.

The first negative is using lower voltage, the 2nd negative is using it tied into a radiator, because the lower voltage doesn't even allow what the peltier can actually do, and the radiator works against the peltier.

With all your experience however you've settled on what works for you, congratulations.

What are you cooling the hot side of the peltier with?

It would be interesting to see how you mounted the peltier to the radiator end cap, that's why we love pictures, they answer questions before they're asked?
 

kermdawg

Distinguished
Apr 16, 2013
48
0
18,530
Hey bud. First of all love your post. Kinda thinkin about planning something like this with my rig, cause the more I look into water cooling the less I like it-If I'm gonna spend 500 dollars on a water cooling setup, I dont wanna have 50-60C temps, I wanna be damn near ambient, like delta T close to ambient.

Wanted to ask you a couple questions on your peltier setup-Firstly, you bailed on the water cooling setup for cooling the hot side pretty quick. Do you think a larger radiator would achieve better results? I know it kind of defeats the purpose in a way.

Secondly- From what I gather on your testing results, you want a real high flow water block with minimal turbulence in the block itself correct? I was thinking that if you had a good solid chunk of copper to absorb some of the cold from the cold side it might be better than say some of these waterblocks with acrylic tops. Basically maybe a 1/4"thick copper base with a 1/4"or 3/8" copper top, soldered/brazed together with a big, hollowed out cavity in the middle, or maybe one of those wide maze setups like the real early water blocks had in em? Basically, something WITHOUT a pin matrix in it?

Thanks for you post bro, you guys are the inovators that give guys like me motivation to try some of this crazy shit ourselves :)
 


It's truly a crappy feeling to spend a chunk of money on water cooling and it fall short of your expectations because the additional heat from overclocking was never fully figured into the equation.

The biggest mistake a water cooler makes using radiators is when he or she has settled on what they think they can get by with when actually they need 2 to 3 times the radiator cooling area to cover their goals.

But they don't actually discover that raw reality until the overclocking begins and the temperatures go up!

Then comes the What Tha?

Followed by that sick feeling in your stomach when your mind begins to total up all you've spent!

Been there done that! :pfff:

So then, I thought there had to be a better way of water cooling, where you had all the control you needed and were not at the mercy of radiators cooled by ambient room air, because if ambient is all you have, it's all you get back.

The Peltier/Tec cooling will easily take you below ambient, it can additionally take you below zero, if you so choose to go that far, but that's not what's being done in this thread.





In the earlier peltier days the peltiers cold side was directly on the CPU die, requiring motherboard moisture insulating, the CPU die was small, the peltier was small, using a much, much, lower voltage.

Those peltier hot sides were water cooled, but the peltiers I am running get way too hot to water cool the hot side, the pump is not designed to run water that hot, but an all metal heat pipe air cooler can take the heat.

A larger radiator makes no difference when it comes to attempting to cool approximately 180f, you cannot put your hand on it without getting burned, my peltiers max temperature is 125c now that is the burn out point, but 25c below that is 100c which is the boiling point of water.

Get the picture!

The peltiers for this type cooling are way more powerful than the peltiers of yesterday, they have a lot more wattage to counter.



The cold pickup water block is very important, both of the blocks I am using have been modified for higher water flow, and the copper base does not need to be thick 1/8" maximum is all you need.

The cold needs to transfer through the base fast and the water needs to pick up the cold fast, because you're using the insulated reservoir to store the cold, and if the cold stays in the block too long it will begin to freeze inside the block.

That's Bad!

Because any ice inside the block restricts the flow, and slows the flow through the block, if enough water freezes inside the water block, to actually block off the flow, the peltier keeps delivering the cold until the ice inside the block expands and splits the block open.

So you modify the flow restriction to zero and block freezing is no longer a problem.

Some ideas regarding the peltier seem extremely logical, but the peltier does not play logical, it has it's own rules.

Discover what it can and cannot do, then use those discoveries to your advantage.

There is a lot of learned the hard way that is not covered in this thread, this thread is sharing what absolutely does work and can be duplicated!

This very setup is running right now while I am typing this, only one peltier is energized, my actual water temperature is 14.8c my room temperature is 22.2c, that's 7.4c below ambient room temperature, CPU idling at 18c, try that with radiator cooling. :)

My graphics cards are radiator cooled, IMO radiator cooling is the best for graphics cooling especially running full coverage GPU water blocks.

Thank You for your kind words! Ryan



 

kermdawg

Distinguished
Apr 16, 2013
48
0
18,530
It's truly a crappy feeling to spend a chunk of money on water cooling and it fall short of your expectations because the additional heat from overclocking was never fully figured into the equation.
The biggest mistake a water cooler makes using radiators is when he or she has settled on what they think they can get by with when actually they need 2 to 3 times the radiator cooling area to cover their goals.

Ya thats the reason I havn't went out and bought my setup yet. Everyone I talk to about it on forums and whatever says "Oh, a 120x1 or a 120x2 will be plenty if your just doing your cpu!". Ya if I want my cpu to be running at 70C all the time. I want my cpu to be runnin at 40-50C while I'm stress testing. Absolute maximum temps.

Hence why I was looking at a giant phobya 160x9 rad(or whatever the car-radiator lookin one they have is.) Not only would I have enough for a cpu, I could put a couple GPUs on there too, and never have to worry. Not be one of these clowns with 3 itty bitty radiators tryin to piecemeal it together cause they lacked the forsight to plan ahead.

But, why bother with all that BS at all if you can just get one of these neat little peltiers? :)

The cold pickup water block is very important, both of the blocks I am using have been modified for higher water flow, and the copper base does not need to be thick 1/8" maximum is all you need.
The cold needs to transfer through the base fast and the water needs to pick up the cold fast, because you're using the insulated reservoir to store the cold, and if the cold stays in the block too long it will begin to freeze inside the block.
That's Bad!

This may seem like a dumbass question, but did you consider running some antifreeze in the water cooling loop? I would think it might give you some leeway in your temperatures because then you wouldn't have to worry about ice forming...

If worse came to worse, you might even run some heat trace on the water block itself to prevent it from freezing...as some sort of backup incase something got out of hand or you forget to turn the fans on or something...The heat trace should keep it just warm enough to prevent ice from forming. It might even be able to help you fine-tune the temperature of the water loop, so you could run it as close to the dew point as possible without having to insulate

I know, it kinda defeats the purpose, but...I dunno.

Also I dunno about some other folks but maybe you could share some of those I-learned-the-hard-way lessons if you don't mind. I know its alot to type up but it would help save a guy like me from making some of the mistakes you already might have made.

edit-Oh! I know another question I was going to ask you. Do you think it would possible to add some sort of resistor into the peltier wiring in order to better control the temperatures?
 
Just snagged one of these to cool my GPUs with, a Watercool MO-RA3, I think the MO stands for monster, it is huge! :)

This may seem like a dumbass question, but did you consider running some antifreeze in the water cooling loop?

No it was not considered, because in my earlier water cooling days I did run anti-freeze and a leak was a freaking mess, there's one learned the hard way, anti-freeze will ruin a wood furniture or floor finish.

Do you think it would possible to add some sort of resistor into the peltier wiring in order to better control the temperatures?

It's possible but not necessary, you can control temperatures by raising and lowering fan speed on the heat sinks cooling the hot side.



 
Status
Not open for further replies.