Thoughts on Refrigeration

jdlech

Honorable
May 31, 2016
168
1
10,715
Refrigeration cooling.... I've been thinking about this.
I know there's two basic problems with using refrigeration: condensation and compressor strength.

Any time you drop a component temperature below ambient temperature, you increase the risk of condensation. Even if you apply the cooling directly to the chip, it can still cause condensation on the back of the chip, or the back of the board. But I think I have a solution.

An actual freezer. Most modern freezers are "self defrosting". The idea behind self defrosting freezers is that a heater element melts the frost from the cooling element. The water is then drained away. But each time frost forms on the element, then melted and drained, leaves less and less humidity in the air inside the freezer. This is also the same general principle which dehumidifiers use. In combination with an air tight design, you can remove virtually all humidity from the air inside a computer case and then run the computer at sub zero temperatures without condensation despite wild temperature differences.

An air tight design would have to be absolutely air tight. The motherboard would have to be completely isolated in a hermetically sealed ice box. I thought of running all cables through the insulated walls to the rest of the case, which would house a second board with all the connectors for peripherals. It would be a freezer inside the computer case housing only the motherboard and whatever is directly mounted to it. All cables would be run though the walls and sealed.

After the motherboard, gpu, memory, and all wiring is installed in the freezer, it would probably have to run several hours to remove the humidity before you could run the computer itself. In addition, it would have to be a pretty strong freezer compressor to handle the heat generated by a motherboard/CPU/GPU/memory. It would have to be more along the lines of an HVAC or automotive A/C compressor, than a common freezer. (a large, bulky device) All else would be outside the freezer, so once installed, you should never have to open the freezer again.

I envision an automotive A/C compressor with a strong electric motor attached. This would probably have to be mounted in a separate box outside the whole computer case with insulated, flexible hoses leading into the freezer box itself. It would also consume a lot of power. But if you're looking to O/C your computer to the levels requiring refrigerant, you're probably not concerned with the electric bill. That would be like buying a supercar while worrying about gas prices.

The noise would be another concern. I don't know how I could reduce noise effectively while maintaining the ventilation a compressor needs. And the heat generated by a strong compressor is enough to heat a small room in the dead of winter. It would need plenty of ventilation, which runs counter to noise reduction. There's a good reason why most HVAC units sit outside the house.

Naturally, the water runoff would be handled the same as a standard refrigerator handles it. Simple containment and evaporation.

The idea of mounting cooling lines directly to chips has the standard problem - you can't do it until you've removed the humidity from the air. There's several ways to deal with this, but I prefer to use a radiator design with a glycol pump that circulates super cooled glycol - with the pump working only after the humidity is removed. This would need humidity sensor that controls the pump and/or valve. A radiator design would have an extra advantage of cooling the chips directly as well as the air and board in general. And glycol can soak up a lot more heat than air - providing more thermal stability even if the system is overwhelmed for short periods of time.

TL;DR
So here's what I'm thinking so far: a hermetically and thermally sealed freezer with a dehumidifier. Kind of a hybrid between an HVAC and a standard refrigerator. All cables sealed through the walls with the PSU and all other components not directly mounted on the motherboard cabled outside the freezer. And a humidity sensor, radiator, and pump providing direct chip cooling as well as a fan to cool the air inside the freezer.

Right now, it's all very abstract. What are your thoughts and concerns? Any better ideas?
 
Solution
The 'PC in a refrigerator' idea has been done numerous times....and each time with less than positive success. Your best bet would be to use the compressor for sub-ambient cooling via chilled liquid. There's even a section of the watercooling sticky that addresses this. There's also a thread that addresses oil immersion cooling.

But hey, you're the expert, let's see what you build and prove us all wrong. :)

Edit: Updated with appropriate links.

http://www.tomshardware.com/forum/id-3038208/overclocking-cooling-water-cooling-sticky-index.html
http://www.tomshardware.com/forum/277130-29-read-first-watercooling-sticky
http://www.tomshardware.com/forum/276288-29-mineral-links...

USAFRet

Titan
Moderator
PC in a fridge or freezer:
Alt title..."What is the fastest way to kill a fridge or freezer?"

The fridge is not made for a permanent heat source inside the box. The duty cycle does not permit that.
It will die an early death.

The other question is..
What problem does this solve?
To keep the components down to some random temp? Why? What specific performance gain does this give?
To be able to OC beyond some random number? Is the OC level an end in itself? Or is the overclock to support some other performance gain?


I've read reviews of people doing this.
Almost all end with..."It was fun as an experiment, but never again"

The noise from this would be substantial.
 

jdlech

Honorable
May 31, 2016
168
1
10,715
:lol:. (is that better?)

If that's an indication of how much consideration you put into a new engineering project, I suspect this thread might not be for you.:lol:
No insult intended. Not everyone enjoys this engineering stuff.

USAFRet
The purpose is to provide significantly better cooling than a standard ambient water cooler can provide by using significantly colder water.
For what reason is up to the person using this system. I want to think on this purely as an intellectual exercise.
And you're right, as I said; there's a good reason why HVAC units are installed outside the house.
Heat and noise might be a total project buster even if all else succeeds.
 

Stubbies

Commendable
Jan 6, 2017
94
0
1,660
Sigh. Not sure why people try to get into these exotic cooling solutions when with all of the money invested in the cooling you could have just bought a better processor in the first place and gone with standard big air or water cooling.

If you are one of those freaks who just has to be able to clock chips super high requiring crazy cooling solutions eh whatever. Freak away.
 

jdlech

Honorable
May 31, 2016
168
1
10,715
Why, thank you Stubbies. Why do people insist on building guitar after guitar after guitar? It presents an interesting intellectual challenge. It's satisfies a desire to make something your own way. Hobbies make people happy.
Mine just happens to be dreaming up better mouse traps, like this PC freezer. To make something work after others gave up on it.
 

USAFRet

Titan
Moderator


"significantly better cooling"
Do current consumer grade components need 'significantly better cooling' ?
Via this construction, you could probably indeed reduce the operating temp.
But does this actually result in higher performance? And performance in what?

Current components are designed to run within the current cooling options. Air, liquid, whatever.
Do you get any better actual performance at 20C vs 65C? Generally, no.

Would you get better performance at -3C? How much, and does that compensate for the hassle, insane power consumption, and noise?
Performance in what? By what benchmark?

If all you're looking for is a screenshot of a functioning PC where the CPU is -3C...go for it.
If you're looking to process large Excel files faster, or gain an extra few FPS in gaming...buy a better system.

Now...if you could get the system running at -100C...then we would probably see a real performance gain.
But you might get about 8 seconds of that.

Like a car engine. We could run it in nitro and LOX. Insane horsepower and speed.
And have to rebuild the engine every evening.
 

RealBeast

Titan
Moderator
Phase change cooling has been around (at the fringes of CPU cooling) for many years but has real issues with condensation and a lack of substantial effectiveness in real world use.

Sure you could do an experimental project with it, but I would go with a chilled oil submersion system first. :lol:

(and no, I would never actually waste my time doing the easier and possibly more practical and painful submersion cooling.)
 

jdlech

Honorable
May 31, 2016
168
1
10,715
Update: I found a compressor that can provide up to 2030 BTUof cooling at 7 deg. C., and is 48Db. That works out to be about 595W. So it could maintain a motherboard at a constant 7 deg C. with up to 595W of heat dissipation. Any heat dissipation beyond that would be unsustainable.

I might be able to reduce the noise level down to 40dba (about the same as a high speed computer fan). But 600W is probably not strong enough for an O/Cd motherboard/GPU combination. Especially with all the thermal isolation of a freezer box.

edit.
Upon further reflection, I can get more than 600w of cooling with ambient air coolers. I would need something on the order of 4x more, which would produce significant noise and consume much more power.
 

jdlech

Honorable
May 31, 2016
168
1
10,715


Fair question: If I can run it stable at 7C. before overclocking, then I have 93 C to go before an intel chip starts to throttle back performance. The idea is to provide MORE cooling than simple ambient temperature water coolers, not necessarily a lower temperature. However, lets say room temperature is 23C, for arguments sake. A refrigerant cooler would provide 16C more "headroom" than ambient air cooler. Keep in mind that all cooling systems are overwhelmed on a local level (the chip), which is why chip temperatures rise until a new equilibrium is obtained (heat production = heat dissipation). The idea is to increase heat dissipation by reducing the ambient temperature by 16C, allowing the user to overclock by whatever amount that raises the chip temperature that extra 16C. It's not about having a "cool" chip, but rather converting that into a faster chip. There's people who spend days and weeks tweaking their computers to eke out a 4% performance improvement. We have competitions based on overclocking.

I would rather have a cooler that dissipates 1800W of heat at 50C than one that dissipates 600W at 7C. Because I could OC a motherboard far more with the former over the latter. But, I still need to consider power and noise too.
 

jdlech

Honorable
May 31, 2016
168
1
10,715
Now I'm thinking another possibility is to augment a standard liquid cooler with an AC compressor - having the compressor cooling the glycol only when the sensors report a higher temperature, say above 60C.

That might be better as it works in addition to the standard heat dissipation of the water cooler.
 

USAFRet

Titan
Moderator
My question is this:

If a particular CPU can run at 100% of rated performance at 45C (typical for a reasonable watercooler at room ambient), and can also run at 100% of rated performance at 7C....what does lowering it to 7C gain you.

Now...if you were overclocking and needed that extra temperature headroom, and are chasing that elusive max OC number just to have that number...then go for it.

But as you stated OC would probably push your design over it's limits (power, noise, reliability)...I'm at a loss to find the usefulness of this.

As an engineering exercise and hobby...go for it.
As a useful PC...meh.


Do it. Build it. Show us.
 

jdlech

Honorable
May 31, 2016
168
1
10,715
Yeah, I just spent a few minutes reviewing the mineral oil aquarium designs. A good advantage is that it uses the whole aquarium surface area for thermal transfer. But as noted by others, heat dissipation is still not enough for extended overclocking. Instead, I think I would weld together an aluminum heat sink aquarium and mount fans on all sides. Quite a lot of expensive aluminum, to be sure, but it would solve the heat dissipation problem.
One guy mentioned dropping a refrigerator condenser into the oil, but the compressor burned out. And maintenance/upgrading would be such a messy affair.

I'm still at the thinking stage with this.
Obviously, any compressor would have to handle sustained operation.

To answer your question, if it can run at 100% at 7C, then you should be able to run it at a sustained 130% at 45C. as the cooling could dissipate an extra 600W of heat. At least, that's the idea.

Which makes me think more towards augmenting a standard liquid cooling system instead of a freezer. It comes with a whole new set of considerations.
 

RealBeast

Titan
Moderator
Good call on passing up oil immersion, total mess to work on.

With any cooling, even with liquid nitrogen, an overclock reaches limits based on other factors that temperature -- specifically things like the voltage applied to the CPU and other components and the stability of other components like the VRMs that would also need significant cooling.

A high quality air or closed liquid cooling solution is far easier to implement and maintain.

You can also do quite well with air if you use a top end air cooler and liquid metal thermal compound. A Kaby Lake i7-7700K running at 5GHz will run at only 80C under full load, which you will rarely experience except during benchmarking, as you can see HERE. Understand that liquid metal needs much more frequent replacement than standard thermal paste though. However, if you are interested in minimizing temperatures at high overclocks it may be helpful.

To me the real question is whether your intended uses and budget justify the additional expense and effort, although there is the hobby component that can weigh in towards more $$.
 

USAFRet

Titan
Moderator
The temp v performance curve is not fully linear.
At some point, other factors start to impinge on the expected performance.

Voltage, temperature of other components, stability.

Increase the voltage and freq, and with a good CLC it is unlikely to hit max temp before other things become the limiting factor.
Not the least of which is the meatsack sitting behind the keyboard.

EDIT: Ninja'd on most of that...lol
 

RealBeast

Titan
Moderator


^ yup this, and as so eloquently stated and missed in my reply "the meatsack sitting behind the keyboard." :lol:
 
IMHO, you're also overlooking a few important points:

Unless you use a sealed, pre-manufactured system (not just a compressor), you will need a refrigerant license and a lot of specialised tools. Do you really want to spend several thousand dollars on stuff you'll probably not use again, plus have a whole pile of liability if you screw up and release refrigerant?

What happens when the compressor cycles? Your computer won't like having ten minutes of no cooling. Any shorter will be an issue with short-cycling, and doing away with the cycling will end up with ice build-up, overcooling, and slugging the compressor (=bad).

 

rubix_1011

Contributing Writer
Moderator
The 'PC in a refrigerator' idea has been done numerous times....and each time with less than positive success. Your best bet would be to use the compressor for sub-ambient cooling via chilled liquid. There's even a section of the watercooling sticky that addresses this. There's also a thread that addresses oil immersion cooling.

But hey, you're the expert, let's see what you build and prove us all wrong. :)

Edit: Updated with appropriate links.

http://www.tomshardware.com/forum/id-3038208/overclocking-cooling-water-cooling-sticky-index.html
http://www.tomshardware.com/forum/277130-29-read-first-watercooling-sticky
http://www.tomshardware.com/forum/276288-29-mineral-links
http://www.tomshardware.com/forum/252012-29-computer-refrigerator-freezer (there's literally dozens of these if you Google for them)
http://www.tomshardware.com/forum/id-2196038/air-cooling-water-cooling-things.html


I can probably find more info, just let me know.
 
Solution