Exploring Below Ambient Water Cooling

Page 22 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
View before insulating

TecBox12.png


There was quite a bit of work from the picture above to this, insulating the tubing is very time consuming!

Looks like a gray snake orgy! ROFL :lol:

TecBox18.png


TecBox19.png


TecBox20.png


Return Lines in Res/Cooler

ResCoolerReturn.png


Output line of Res/Cooler

ResCoolerOutput.png


Line to pumps

FeedLinetopumps1.png


Feedlinetopumps2.png


I'm running exactly 1 gallon of water in the reservoir, the feed line has a Y branch splitting to 2 lines each line feeds a pump intake, one pump cools the CPU loop, the other pump cools the GPU loop.

Each independent loop passes through it's own peltier cooling before returning to the insulated cooler/reservoir.

 
I've already made a major change and a confirmed discovery, the GPU peltier using the full copper block was stalling out!

Why?

Simply because of the copper mass of the water block used retains too much cold, the discovery is the cold side cannot amass too much cold or it chills the hot side pf the peltier, the most optimized is to have a cold water temperature pickup water block as close to the actual peltier dimension as possible.

So I went back to the modified flow XSPC Rasa water block for the GPUs peltier cooling loop, it's base is exactly the same size of the peltier itself as the pictures show.

GPUpeltier2.png


GPUpeltier1.png


GPUpeltier3.png


GPUpeltier4.png


I'm using automotive trailer light connectors in the peltier wiring

TrailerLightsconnectorforpeltier.png


Better CPU peltier pictures

CPUpeltier1.png


CPUpeltier2.png


CPUpeltier3.png


I'm presently running 7c below ambient room temperature and have not done any airflow fine tuning as yet.
 


Seeing as how this all started as an experiment not knowing the outcome at all.

Efficiency would be seriously debated on the negative side as there's a lot of fine tuning experimenting involved, juggling the hot/cold peculiarities of the peltier, to get the maximum cooling from it.

Plus the additional electrical load to run this peltier cooling in the first place, it adds an additional 500w load above my operational computer.

However it is a dependable below ambient cooling solution as since it's incorporation neither my CPU or GPUs have experienced even ambient room temperature, they've been below ambient the entire time.
 


To save build time!

This is a continued experiment.

The CPU peltier was already in service and just required tubing rerouting after the radiators were removed.

Since it is an experiment, I have not closed any doors or burnt any bridges, and mounting it all inside or outside a case will be a challenge.

Because the peltier coolers exhaust heat and they have to be kept away from each other, or they'll cancel each other out.

The perfect internal mounting solution for this type cooling doesn't exist, I've already considered buying a metal break and making an enclosure from scratch.

An all in one solution would be nice but even the mountain mod cases are not designed to handle this cooling, so the end results will probably be an external enclosure, however before I break out the metal sheers and invest in a metal break, the experimenting will continue at least for a while.

It all needs to be time tested, that's the unknown factor?


 
Testing conclusions at this time:

Graphics cards IE. 2 580GTX in SLI cooled by full coverage water blocks, produce too much heat to be added as a single loop tied to an overclocked CPU, temperature results running 2 peltier coolers were only a few degrees below ambient under gaming load.

I ran a 2nd test rerouting the tubing and giving the GPUs their own insulated reservoir to see what would happen regarding the peltiers cooling capability vs gaming temperature, (No pictures covering this experiment.), the peltier was able to produce impressively low SLI GPU idle temperatures, unfortunately could not maintain those temperature ratio levels (idle/load), under gaming load.

Once the temperature increases in the Res/Cooler it takes time, too much time, for the peltier to recover and drop the water temperature back down to the original testing temperature.

Peltier cooling could probably handle 1 GPU but not 2 and I have no intentions to remove one of the GPUs to test that, simply because the facts I already have from all my testing back my conclusion below.

So my conclusion at this point, is the absolutely best way to cool full coverage water block multiple GPUs is traditional closed loop water cooling using radiators, the radiators will allow the load temperatures to reach a certain level and hold the temperature at that load level, the peltier cooling solution will not hold the temperature. Multiple GPUs under gaming load produce too much heat for the peltier cooling to dissipate fast enough.

Peltier cooling is far better suited to control below ambient cooling with an overclocked CPU, use radiators to cool the GPUs!

Now my further intentions and testing plans are to return the GPUs to radiator cooling and further experiment with the 2nd peltier added to the CPU loop.

Reworking and rerouting everything will probably take a whole day to do, so I'll share more information when I have it to share.

Best to All!
Ryan
 
This is a very interesting result. I would imagine with the addition of 1-2 TECs in the loop, you still might be able to run subambient with the 2 GPUs, but it might depend on the total configuration plus you have to consider the additional power draw to run the additional peltier units. I think most people would agree that running 40C lower than air load temps on GPUs would be just fine for them with a normal water loop (raises hand slowly...), but in the quest for sub-ambient temps, I know it wasn't exactly what you were wanting.

Very interesting findings though.

Have you thought about running a chiller loop with TEC units inside your res as a cooling loop for the reservoir volume? I'm thinking a snaking loop of copper or even a cheap radiator that would interface like a reverse slush box and would run cooled water/coolant through it simply to cool the reservoir water? Your thermal foam blanket would be useful in this manner and by keeping the loops separate, you might be able to continually keep the 'cooler' loop colder and transfer more to the reservoir body. I'm thinking along the lines of the cooling coils of a freezer or refrigerator as an example.

Of course this would be adding another TEC or 2 and at that point, you might just be better off adding them to your normal configuration to cool the GPUs..

chill_box.jpg
 
Spent an entire day on this setup just fired it up at 11:30P, the GPUs are now being cooled by a 360 and 240 rads on an independent loop, the CPU is now being cooled by 2 peltiers.

When I fired up at 11:30P the water temperature in the cooler was 24.7c at 12:00A, 30 minutes later the water temperature in the cooler was 14.2c, a 10.5c drop in 30 minutes.

This is the most drastic temperature drops I've had yet, 2 peltiers are really cranking out the cold, there's presently only 1 gallon of water in the cooler.

I'm posting this at 12:13A and the water temperature is 11.5c thats a 13.2c drop since 11.30.
Ryan
 


Continued testing.

I added the 2nd TEC to see if it could hold the GPUs added heat and it could not, it may have been possible with a GPU hybrid setup meaning only the GPU cores being water cooled and the Ram and VRs air cooled, but I'm running full coverage water blocks which means the heat from the Ram chips and voltage regulation are also added to the mix.

Now if I was only running 1 full coverage GPU then 2 TECs could handle that load temperature, but it cannot handle 2 GPUs in the same cooling loop with the CPU.

I'm presently running 2 TECs on the CPU loop, the GPUs are now being cooled by radiators in a traditional closed loop setup.

If you had the 2 TECs already in your possession wouldn't you continue testing?

It's a level of curiosity I have to have answers to, for myself if no one else cares.

I'm after a viable, duplicateable, dependable, below ambient cooling solution and in the beginning I wasn't even cooling any GPUs it was all about cooling the CPU and was using ice to do it.

The TEC/Peltier cooling was an attempt to get away from the load and burden of freezing and changing out the ice every day.

The TEC cooling is actually cheaper than running 2 freezers every day running at increased load to freeze the ice, so electrically it is acceptable to me and cheaper than what I was already doing.

From the very beginning this has been about discovery, and sharing these discoveries with those that may be frustrated at the cooling they're presently running, because it is just not performing as they expected it too when they bought it.

This does not apply to someone totally satisfied with what they are running.
Ryan
 


Most at other websites don't share what they've learned and they don't want you to know their secrets, I'm doing this as a THGF open reference to others that may be curious of what could be accomplished using this type of cooling.

Will others do it?

I don't know, well you're now dabbling into it, so you're curious as well! 😉

😱 Just dropped below 9.0c I need to disconnect one of the TECs, or I'll be into condensation territory!

Note to self: Install a switch on the 2nd TEC!
Ryan
 
Discoveries regarding 2 TEC cooling of the CPU;

Even 2 TECs cannot maintain water temperature in the cooler under CPU stress testing, Prime95 and Intel Burn will begin to heat the water in the cooler approximately .8c in 30 minutes.

Gaming load cannot override even 1 TEC as the water temperature in the cooler continues to drop even gaming.

I increased my water mass from 1 gallon to 1 1/2 gallon because of how fast 2 TECs drop the water temperature taking me below 10c, however 2 TECs are closer to what was being achieved using the ice.

I'm presently at 10c running 1 TEC, as I had to shut down the 2nd as I was going below 9c, the 1 TEC by itself will maintain around a 13c range, the water temperature in the cooler is increasing as I type this it is now at 10.3c.

Water temperature in the Cooler/Reservoir cannot be allowed to drop below 8c, that's the point condensation begins to form as I have not taken any motherboard insulating steps, so I operate in a range above 8c.

At this point I can safely surmise that 3 TECs would be capable of taking it to sub zero temperature levels so fast it would blow your mind!

That would require adding anti-freezing solutions to the water, and fully insulating the motherboard as well!

I have no intentions of doing that, but it is capable of it.

Presently up to 11c on 1 TEC.

Regarding the GPUs now cooled by closed loop radiators their maximum load temperature reached under maxxed out benchmark load running multiple loops of the benchmark program yielded a 41c load that immediately recovered to a 33c state after the testing was stopped.

The TEC cooling could not recover that fast, cooling 2 GPUs, 41c is 39c below the 80c load of the stock air cooling of the graphics cards.
Ryan
 
How about........

blocking the hotside of one Tec and looping it into the res to remove the Tecs heatload? could that maybe offer a balance load to keep it dropping to 8'c or below?
I know you had issues nearly melting a pump before but that was slightly different iirc,
this would have the flow from the Tec going into the mass and the pump would pull from the mass thereby avoiding a massive heatstrain on the pump/seals as the fluid passing through the pump would be cooler than previously?
i reckon with tweaking mass you could finetune that to a stable temp margin relatively easily
worth a thunk maybe?
Moto
 


If I just use 1 TEC it doesn't happen, the TECs are water looped in series, the water flows through the first then through the second then dumps into the R/C (IE. Reservoir/Cooler for future reference), the simplest solution besides what you're suggesting is just unplug the 2nd TEC.

I'll more than likely just put a switch in the line but it has to be a switch that can handle 26amps @ 12v, that will give me quick temperature drop to a certain temperature level then switch off the 2nd TEC to run at that temperature level.

Increasing the water level mass doesn't slow the temperature drop that much, only very slightly and it will still drop below 8c if both TECs are active, so water mass in the R/C isn't the solution.

I have yet to play around with hot side manipulation to see what control that may give, so there's still testing to do.
Ryan
 
I was thinking that the increased water mass would take longer to both heat up and cool, and obviously the cooling of the tecs would negate the heat input and hopefully balance things in a controllable manner,
the switch may be the way but that would require more monitoring than the heating/cooling simultaneously option, just trying to make it a simpler solution for you :)
Moto
 
Since I'm using a smaller cooler now I can only increase thermal mass to 2 gallons and another half gallon and I'm there, but you are right more mass takes longer to cool.

I've already installed the switch, when I first boot up I'll use both TECs to quickly drop the temperature of the water mass once it gets down to 13c, I'll cut off the 2nd TEC and the first TEC by itself can maintain the temperature in the 13c ~ 15c range, even 15c is 9c below ambient.

With both TECs energized it will drop the water temperature approximately 10c in 30 minutes, that's some serious cooling! :)
 
Just dropped below 9.0c I need to disconnect one of the TECs, or I'll be into condensation territory!
Click to expand...

This is why I was asking why two TECs. I understand bringing in the second when you attempted to cool two GPUs. But when you switched them out of the loop, I would have just removed the other TEC.
 


Two TECs drop the temperature of the thermal mass of the water to the operational level I want to run it at, twice as fast.

Once that operational temperature level is reached I switch off the 2nd TEC and let the 1st TEC maintain that level.

One TEC by itself has only been capable to take the temperature level in the cooler down to 12c, using 2 TECs I can drop the level as far as I safely desire to run it for overclocking testing purposes, (Meaning above the condensation forming temperature range.).

When I started up the machine this morning my ambient room temperature was 23.4c the water temperature in the cooler was 20.6c.

With both TECs running @ 15 minutes the water temp was 17.3c, at the 30 minute mark 14.6c, at the 45 minute mark 12.6c, 1 hour 10c, so I just cut out the 2nd when the operational level I want to run at is reached.

The 2nd peltier with 1 1/2 gallons of water drops the operational temperature level of the cooler faster than 2 blocks of ice did running 8 gallons of water.

This is all about discoveries. Ryan
 


Moto, The switch is working great!

It has given me amazing options! :)
 
Excellent news mate :)
now its just a case of settling in and getting used to the performance as you see fit/need
Good job man, now you just need to make it all pretty-like in a box :)
Can I have a development job when you go on the market with this? 😛
Moto
 
There's still so much to learn from this cooling, for example your Chillbox Cooler uses traditional finned heat sinks for the hot and cold side transfer.

So is it even necessary to use the heat pipe air coolers I'm using or could it be done with much cheaper heat sinks?

Would a non modified water block be able to be used so as not to destroy a water blocks warranty to run this cooling?

There are so many unanswered questions I still have to find answers to?

By the time it could reach pretty in a box status, I'm sure some manufacturer would have already stolen it and claimed it as their own idea.

A lot of my work has already been stolen from THGF anyway!

One website stole the Thermal Compound Roundup and posted at their site and another website has stolen my Black Edition Overclock Guide!

Chris Angelini handled the Thermal Roundup theft and had it removed from the perpetrating website, because it is Toms Hardware and Best of Media property.

I guess I should be flattered they thought it worthy enough to steal, but I invested a lot of work in those guides.

http://forumhardware.net/forum/overclocking/amd-black-edition-overclock-guide/

But this thread is open to all, sharing all my discoveries, so anyone could duplicate the cooling, if they had the fortitude and determination to do so, and not be at the mercy of some manufacturers inflated pricing!
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom