Anybody have a liquid coolant leak in their PC and did it go well after regarding the cleanup?

[Berwhale]

...T4 is the temperature of the water at the CPU while T5 is the temperature of the water at the radiator; we can take T5 to be equal to T4 with negligible error

If T4 and T5 are the same, then what is the radiator doing?

 

[JustAnandaDourEyedDude]

If T4 and T5 are the same, then what is the radiator doing?

Quite right, they are not the same, and I am being imprecise, thanks for pointing it out. There are differences among all of those temperatures, and T1 > T2, T2 > T3, T2 > T4, T5 > T3 and T5 > T4. The water temperature is raised from T4 to T5 in going through the CPU water block and drops back from T5 to T4 in going through the radiator. (q1=) mdot*cp*(T5-T4) is the bulk heat heat change for the water side at both the CPU water block and the radiator. There is also a water convection coefficient and temperature drop on the water side of the radiator, which I omitted on purpose.

However, when I said "we can take T5 to be equal to T4 with negligible error", I am looking at how well we could cool the CPU, and I am taking all but one (T1-T2) of these differences to be small relative to the difference T1-T2. My intent is to point out that with good air- and water-cooling we are already in the situation where the driving temperature differences for the various convective cooling sites are small relative to T1-T2 and we have already brought T2 down most of the way to T3. Taking the cooling to the limit by pushing velocities, mass flow rates and heat transfer contact areas to the limit only gets us to the limit situation where T2=T4=T5=T3. The ambient T3 sets the limit on how low T2 can be and thus limits how low T1 can be brought.

When the water flow rate is bumped up and the contact areas at the radiator and CPU block are bumped up, the difference T5-T4 will decrease correspondingly. I will edit my original post to clarify.

 

[Berwhale]

The water temperature is raised from T4 to T5 in going through the CPU water block and drops back from T5 to T4 in going through the radiator

I think you have T4 and T5 the wrong way around.

 

[JJB70]

Heat management is an important subject and it's a good example of the old adage that there's more than one way to skin a cat. Really the choice between liquid and air cooling is generally about space. Liquid cooling is more efficient but you still need somewhere to reject heat from the coolant (eg. fan cooled elements/radiators, sea or river water coolers) and if the space is there then air cooling can work perfectly well depending on what temperature is required. If necessary you can use chillers to bring air temperature down, you may also use chillers with liquid cooling if heat rejection to air or sea is inadequate for the application). I have worked in power plants using air cooled condensers for high pressure Rankine steam cycle plant and they worked well, but they were enormous compared to an equivalent water cooled condenser. I have also worked in plants using chillers on gas turbine air intakes to boost performance.

 

[Berwhale]

The Corsair H60 in my current PC cost £65, or to put it another way, 5.3% of my overall system cost. I could have saved maybe £25 by buying a decent air cooler. However an air cooler takes up more room in the case (I have a small mATX one) and offer less opportunities for optimizing airflow - my radiator exhausts directly out the back of the chassis and I could move it to the front or top of the chassis if I wished - I guess you could try to do the same with ducting on a air cooler, good luck with that!

 

[JustAnandaDourEyedDude]

I think you have T4 and T5 the wrong way around.

The convective heat transfer coefficient is typically based on taking the driving temperature difference potential to be the difference between the solid wall surface temperature and the bulk or "infinity" or "free-stream" temperature of the incoming fluid. Although some authors use a "film" temperature when preparing their convective coefficient correlation; you have to examine the provenance of the particular coefficient correlation. Consistent with that, at the CPU water block, the convective heat transfer coefficient h4 is calculated based on the incoming water temperature T4. After diffusion of heat across the cross-section, the bulk water temperature is raised to a higher value T5, which is then the bulk temperature of the incoming water at the radiator, upon which the heat transfer coefficient h5 is based. After passage through the radiator, the bulk water temperature drops back to T4 for return to the CPU water block. So I think I have T4 and T5 the right way around. There is also a small rise in water temperature in passing through the pump (the pump work raises the stagnation enthalpy of the water, which enables it to overcome the frictional forces in the circuit), but I am neglecting such details in the interest of seeing the big picture.

 

[Berwhale]

but I am neglecting such details in the interest of seeing the big picture.

The 'big picuture' includes the case and all the other components and cables within it. The CPU + cooler is not a closed system.

 

[JustAnandaDourEyedDude]

The 'big picuture' includes the case and all the other components and cables within it. The CPU + cooler is not a closed system.

True enough, but I am just interested in seeing the big picture that using only convective cooling, even when done to perfection, the best you can hope for is to lower the CPU surface temperature T2 to the available sink temperature T3 in the extreme limit. And with the better air-coolers taking us most of the way there, there is limited scope for improvement, which explains the trend of why water-cooling is not much more effective than the best air-cooling. For years, there were legions of people, particularly gamers, paying oodles of money for water-cooling probably without realizing why water-cooling was limited by the same ambient sink temperature that heat is ultimately rejected to; though water-cooling has gotten cheaper now.

Sure, one can argue that with a congested case and lots of other hot components in reality, water cooling may have an advantage that it has a lower available sink temperature outside the case, while air cooling has to deal with the higher sink temperature available inside the case. As I mentioned in my original post, a benefit of water-cooling is in being able to increase the compute-density. But then one could argue back that the limit case of convective cooling "done to perfection" includes perfectly cooling those other components by driving a very large air-flow through the case and routing it properly to bring all surfaces to the ambient temperature T3 outside the case. For consumers who want better cooling than the stock cooler that comes with the CPU, I am assuming that good air-cooling includes well-designed uncongested cases with adequate ventilation.

 

[Chromatischism]

Usually when the rubber hits the road and overclocking tests are done, water achieves stable CPUs at higher clocks because of its stability. Air allows more micro fluctuations in temperatures that lead to the system locking up or rebooting.

 

[NTK]

TL;DR : Water-cooling costs more, is noisier, does not lower the max CPU temp by much more than what the best air-coolers achieve, while introducing some small risk of catastrophic electronics failure due to liquid leakage. The diminishing returns are due to the limitation that the lowest CPU surface temperature that these convective cooling methods can produce is that of the ambient air temperature. This limitation could be overcome by using a refrigeration cycle device instead of liquid-cooling. ...

Very interesting analysis. I haven't looked into this since I don't have any sort of high performance PC, so it is not something that I am concerned with. However, a significant portion of my professional work is in thermal analysis/management, and why not take this opportunity and have my own look at the PC cooling system.

In a high level view, after breaking down the system into its sub-components/sections, the heat flow path is (with many of the finer details left out):

Air cool:

CPU die → internal thermal interface material (TIM) → integrated heat spreader → cooler baseplate → heat pipes → "radiator"/fan → ambient air​

Liquid cool:

CPU die → internal thermal interface material (TIM) → integrated heat spreader → cooler baseplate → cooling liquid → "radiator"/fan → ambient air​

So basically the primary difference between the 2 systems is air cooling uses heat pipes to move the heat from the baseplate to the radiator fins, and liquid cooling uses liquid transport. Conceptually, the other items can be made equivalent, e.g. cooler baseplate, radiator size, fan speed/flow rate, etc. Therefore, it is mostly down to the difference between the heat transfer effectiveness of heat pipes vs liquid transport.

Heat pipe cross-section (from ACT):

Heat pipes work optimally at their design temperature range. The way a heat pipe works is for the liquid inside it to vaporize and absorb the heat at the heat-input end (hot end). The vapor flows to the heat-output end (cold end) due to the pressure differential (pressure goes up at the hot end due to liquid vaporizing, and goes down at the cold end when the vapor condenses). The vapor condenses at the cold end and releases the heat, thereby transporting the heat from the hot end to the cold end. The condensed liquid then flows back to the heat-input end by capillary action and gravity assist, if the hot end is at a lower elevation than the cold end. (Heat pipes perform better with the cold end on top and hot end at the bottom.) Because a heat pipe works poorer at a lower temperature than at its optimal operating temperature, it tends to move the system toward its optimal performance temperature. The liquid boiling temperature is adjusted by the choice of liquid and the pressure inside the heat pipe.

Moving heat with liquid transport is conceptually much simpler, and the liquid cooler works equally well throughout a much wider temperature range. The temperature difference between the hot and cold ends will also be minimal (and less than heat pipes), meaning more effective heat transfer. So I do think liquid coolers should have some performance advantages over air coolers.

An example performance heat pipe performance graph using the ACT heat pipe calculator:

Free Heat Pipe Calculator

Quickly calculate thermal performance and design parameters for heat pipes using ACT’s free online Heat Pipe Calculator – ideal for engineers and system designers.

www.1-act.com

 

[Doodski]

Very interesting analysis. I haven't looked into this since I don't have any sort of high performance PC, so it is not something that I am concerned with. However, a significant portion of my professional work is in thermal analysis/management, and why not take this opportunity and have my own look at the PC cooling system.

In a high level view, after breaking down the system into its sub-components/sections, the heat flow path is (with many of the finer details left out):

Air cool:

CPU die → internal thermal interface material (TIM) → integrated heat spreader → cooler baseplate → heat pipes → "radiator"/fan → ambient air​

Liquid cool:

CPU die → internal thermal interface material (TIM) → integrated heat spreader → cooler baseplate → cooling liquid → "radiator"/fan → ambient air​

So basically the primary difference between the 2 systems is air cooling uses heat pipes to move the heat from the baseplate to the radiator fins, and liquid cooling uses liquid transport. Conceptually, the other items can be made equivalent, e.g. cooler baseplate, radiator size, fan speed/flow rate, etc. Therefore, it is mostly down to the difference between the heat transfer effectiveness of heat pipes vs liquid transport.

Heat pipe cross-section (from ACT):

View attachment 163981

Heat pipes work optimally at their design temperature range. The way a heat pipe works is for the liquid inside it to vaporize and absorb the heat at the heat-input end (hot end). The vapor flows to the heat-output end (cold end) due to the pressure differential (pressure goes up at the hot end due to liquid vaporizing, and goes down at the cold end when the vapor condenses). The vapor condenses at the cold end and releases the heat, thereby transporting the heat from the hot end to the cold end. The condensed liquid then flows back to the heat-input end by capillary action and gravity assist, if the hot end is at a lower elevation than the cold end. (Heat pipes perform better with the cold end on top and hot end at the bottom.) Because a heat pipe works poorer at a lower temperature than at its optimal operating temperature, it tends to move the system toward its optimal performance temperature. The liquid boiling temperature is adjusted by the choice of liquid and the pressure inside the heat pipe.

Moving heat with liquid transport is conceptually much simpler, and the liquid cooler works equally well throughout a much wider temperature range. The temperature difference between the hot and cold ends will also be minimal (and less than heat pipes), meaning more effective heat transfer. So I do think liquid coolers should have some performance advantages over air coolers.

An example performance heat pipe performance graph using the ACT heat pipe calculator:

Free Heat Pipe Calculator

Quickly calculate thermal performance and design parameters for heat pipes using ACT’s free online Heat Pipe Calculator – ideal for engineers and system designers.

www.1-act.com

View attachment 163982

From what I read some time ago the water cooler is better at removing heat faster when the processor load is reduced or as the processor load increases and decreases during normal operation. Where a heat pipe air cooler system needs more time to conduct the heat to the air cooler fins due to thermal resistance in the metal and due to the heat mass in the cooler causing a decrease in the temperature differential between the CPU and the heat pipe causing less thermal conduction. I noticed when torture testing my AIO cooler and processor that the temps when the processor load is decreased drop very rapidly as in maybe 2-3 seconds at most. When observing a heat pipe air cooler I saw the temp drop over a much longer period of time. Not sure how this relates to overclocking a PC if the max temp heat dissipation @100% load is all one is concerned with.

 

[NTK]

From what I read some time ago the water cooler is better at removing heat faster when the processor load is reduced or as the processor load increases and decreases during normal operation. Where a heat pipe air cooler system needs more time to conduct the heat to the air cooler fins due to thermal resistance in the metal and due to the heat mass in the cooler causing a decrease in the temperature differential between the CPU and the heat pipe causing less thermal conduction. I noticed when torture testing my AIO cooler and processor that the temps when the processor load is decreased drop very rapidly as in maybe 2-3 seconds at most. When observing a heat pipe air cooler I saw the temp drop over a much longer period of time. Not sure how this relates to overclocking a PC if the max temp heat dissipation @100% load is all one is concerned with.

My hypothesis is that the condensed liquid in the heat pipes is not moving fast enough from the cold end back to the hot end. Imagining a scenario that the liquid at the hot end is all vaporized (hot end is dried), with the condensates all stuck at the cold end, dependent on capillary action to migrate back to the hot end.

Wonder if you can run an experiment. Position your computer with the cooler radiator on top and the CPU at the bottom, run your test. Flip the computer upside down (so that the CPU is now on top) and test again. See if there is any difference. Theory is that the CPU at the bottom configuration should perform best because gravity assist will help the liquid move from the cold end to the hot end.

 

[Doodski]

Wonder if you can run an experiment. Position your computer with the cooler radiator on top and the CPU at the bottom, run your test. Flip the computer upside down (so that the CPU is now on top) and test again. See if there is any difference. Theory is that the CPU at the bottom configuration should perform best because gravity assist will help the liquid move from the cold end to the hot end.

I read that if there is air in the system and the pump in the heatsink mounted on the processor has air run through it can damage the pump. So I don't want too flip my PC upside down. When the last pump failed both of the hoses where pretty hot near the heatsink and the heat was conducting up the hose. That's how it was able to operate with web browsing and average daily use. The pump is in the water block on the processor and with the present configuration gas cannot get to the pump.

 

[NTK]

I read that if there is air in the system and the pump in the heatsink mounted on the processor has air run through it can damage the pump. So I don't want too flip my PC upside down. When the last pump failed both of the hoses where pretty hot near the heatsink and the heat was conducting up the hose. That's how it was able to operate with web browsing and average daily use. The pump is in the water block on the processor and with the present configuration gas cannot get to the pump.

Sorry. I didn't make myself clear enough. The test I proposed was only for air cooled system (that relies on heat pipes to move the heat from the CPU to the radiator), not water cooled. Water cooled systems shouldn't be affected by the mounting orientation, except for the practical aspects that you described.

 

[Doodski]

Sorry. I didn't make myself clear enough. The test I proposed was only for air cooled system (that relies on heat pipes to move the heat from the CPU to the radiator), not water cooled. Water cooled systems shouldn't be affected by the mounting orientation, except for the practical aspects that you described.

Hehe. I was going to do the test and then I thought nope I just received the AIO cooler back from Corsair California warranty service depot and I don't want to have to send in another so soon. So you almost got test results. That and I realized that if there is no air in the loop and the pump is cycling coolant then the orientation should not matter what the orientation is. Additionally the air is forced up out the top of the case and laying it down on the top would block the vents. I read too much into your comment. Going on 6 hours of sleep last night so I'm not in tip top condition.

 

[Tks]

These newer specialty dielectric fluids are quite a leap from the typical mineral oil immersion in terms of longevity and other characteristics. They can easily last 5-10 years or longer and come with warranty for that long if desired. Performance wise, it may not be that much greater than a typical high performance water/glycol cooling loop but it offers lots of distinct benefits on top of marginally better thermals, namely cost savings. For example, I plan on always buying at least 1-2 of the latest/greatest GPU available every generation, and spending money on water blocks is not necessary with an immersion setup. Buying waterblocks and a few sets of good thermal pads over the life of the fluid will easily cost $2000+ for a couple pairs of gpus over that time period. There may be unforseen CPU socket changes which now needs a new CPU water block which further adds more to the cost. The fittings and hosing/piping for all those blocks also have a cost, and blocks should require semi-regular maintenance which also takes time and effort, coolant swaps etc.

Another huge advantage is that immersing in this fluid allows for subambient cooling with zero risk of condensation. Subambient cooling is already quite inefficient and can rarely be done unattended or without elaborate insulated and controlled setups, but this makes it effortless and viable for continuous 24/7 operation. So if I opt for this I further increase performance and component longevity.

A large tank isn't really necessary- it only needs to be large enough to submerge the motherboard and gpu(s) so for my specific setup with traditionally horizontally mounted gpus is only 15L (10" x 12" x 8") which is smaller than most small form factor cases. If one only has one GPU and mounts it vertically, it can be done with nearly half that volume- 7.5L. You can even fill in the void space with things like pieces of acrylic to use even less volume of fluid. 20L of this fluid ranges from about $300-600 depending on quality/warranty option, typically about $400-500. Other than the one off cost of building a small custom tank/case, this is already significantly cheaper than buying a number of water blocks.

Also regarding cleanup- it is an extra step but it is trivial all things considered. I used to be into LN2/extreme overclocking so I am no stranger to motherboards coated in vaseline and craft putty. Dunking a motherboard or gpu in an ultrasonic cleaner with some isopropyl alcohol and leaving it for an hour or so basically has it coming out newer than it looked coming out of the retail box. Mineral oil is simply too volatile and takes up ions too easily/poor materials compatibility which is what causes so many problems for people. These diaelectric fluids also have a specialised solvent for them which costs $100 (and can be reused multiple times) that you can wash all the PCBs with and leave to air/blow dry and there is no way of knowing it ever touched this fluid. Second hand sales are not a problem at all, and the components basically look like they have never been used if you do the process properly.

Honestly, the biggest issue with this is that component swaps are a bit more involved to do them properly because of the cleanup. This also scales extremely well. If I wanted to add more gpus lets say, I would at most have to add maybe another 0.5-1 L of fluid each, and in many cases I would have to use even less fluid per component as the component itself displaces volume, depending on the configuration of the vessel.

If you’re buying the latest and greatest hardware, size and thermals are a problem, allowing the liquid to naturally rise and sink isn’t enough especially if you’re going to be running systems like a 12900K with something like a 3090. You’re looking at easily over 500W loads. Forget sub-ambient, there no way you’re going to achieve that passively with a top spec system running gaming or rendering duty. And if it’s not passive, I don’t know enough about having active cooling of the liquid (running loops to cool it?).

 

[jae]

If you’re buying the latest and greatest hardware, size and thermals are a problem, allowing the liquid to naturally rise and sink isn’t enough especially if you’re going to be running systems like a 12900K with something like a 3090. You’re looking at easily over 500W loads. Forget sub-ambient, there no way you’re going to achieve that passively with a top spec system running gaming or rendering duty. And if it’s not passive, I don’t know enough about having active cooling of the liquid (running loops to cool it?).

Oh, it's definitely not passive. The continuous load is at least 700W when "idle" (cryptomining when they aren't being used) and even higher when it is being used on other tasks. The tank itself will be connected to an external radiator setup. Specifically, there will be a pump circulating the heat from the top of the tank on one side to the bottom of the tank on the other, and in between this line will be a small brazed heat plate exchanger.

Then, connected to this heat plate exchanger will be a typical external radiator setup with another pump. The rad will have 200mm fans running at low rpm, which is virtually inaudible. I will probably be using two of these even if I don't do this immersion project because I like the silence regardless and can use it with my existing loop:

Its possible to run the dielectric fluid directly from the tank to the radiator like a typical WC loop, however this is a lot of unnecessary volume of dielectric fluid (which is expensive). So a plate exchanger in this situation is ideal so you can use cheap water/glycol coolant on the air exchange while keeping the good stuff in the tank.

In the winter, the 700W+ load will be some very nice, silent, passive heat for the home. In the summer, the versatility of the external radiator can allow me to dump the heat outside without heating up my space further. I've seen people do interesting things like use the waste heat from their setups to do things like heat water in their homes or for under-floor heating, or even using running water from indoor tap/plumbing to cool the system with a heat exchanger in places where water is cheap or virtually free. I know a guy that personally dumps all his heat from a 30KW immersion setup into a nearly lake on his property and circulates the cool water from the lake back up to his shed. It can be scaled up pretty easily and the heat exchanger can be as small as a PC radiator with fans to an industrial dry cooler.

The setup I described is really not much different from traditional watercooling. It's just that instead of having multiple waterblocks and connections between components, everything is just submerged in tank and the only connections are between the tank and however you're exchanging the heat.

 

[Berwhale]

Negative pressure liquid cooling

More about it here:

And here:

Data Center Cooling Technology | Liquid Cooling Solutions

An efficient direct-to-chip liquid cooling system that reduces data center operating costs and eliminates the risks associated with liquid cooling.

www.chilldyne.com

 

[Willem]

I appreciate that the most powerful computers will need cooling for a long time to come. However, with increasing energy efficiency of computers fewer and fewer users will need that. I recently replaced my ageing desktop computer with a modern NUC i7 10th generation (with plenty of memory and a big SSD), and fitted it into an Akasa Turing FX fanless case. The computer is pretty fast, and with this fanless case it is totally silent. I waited a few years until fanless computers would be fast enough for my use case. I am sure that fanless will be a viable option for and more more intensive applications, and sooner rather than later. Energy efficiency is an increasingly important design target.

 

[Chromatischism]

I appreciate that the most powerful computers will need cooling for a long time to come. However, with increasing energy efficiency of computers fewer and fewer users will need that. I recently replaced my ageing desktop computer with a modern NUC i7 10th generation (with plenty of memory and a big SSD), and fitted it into an Akasa Turing FX fanless case. The computer is pretty fast, and with this fanless case it is totally silent. I waited a few years until fanless computers would be fast enough for my use case. I am sure that fanless will be a viable option for and more more intensive applications, and sooner rather than later. Energy efficiency is an increasingly important design target.

Efficiency is often used up to increase performance of processors. That creates more heat. Yours was designed with a different goal.