Cooling research


Feb 3, 2020
Hey guys !

I am a student at the University of Technology in Delft.
Together with a group of fellow students and researchers we are investigating a new concept of liquid cooling which seems be very useful for the cooling of ICT components and computers.
It works pretty much the same as water cooling, using a fluid for cooling, thus using a radiator and fans to transfer the heat out of the fluid.
The main difference is that you don't need a pump. The fluid runs using a specific physics phenomenon (I can't really go in detail here, sorry about that .. ) and temperature difference.
This means, the higher the temperature difference, the higher the flow of fluid.
And although we cannot reveal much about the actual working, its not capillary action or phase change (as used in heat pipes).
It does however require a special fluid to operate, which is a little more expensive than destilled water or any common coolants. Yet, we estimate that it will not be overly expensive (lets say over €50,- per litre).
So, no pump, (nearly) no failure and less noise ...
But still, sounds good right ?!?

We don't really have a prototype or anything yet, it's just theory and math that says that all of this should work.
Therefore, I can't say anything yet about pricing, we just don't know yet.

Now, here's the thing. We think the technology is up to it, but we are looking to get better insights into what properties make a good cooling system and what properties you value when buying a cooling system. What do you think a cooler should feature? What features should a cooler have? What would you want a product to be?
Furthermore, we are interested to see if you guys think it'll add something to the market or would you recon there is already enough products out there?

BTW, as we are researchers, we would like to use your replies in our research. We are not sure yet how we are considering doing this, but we will somehow attribute your statements to you of course (using your nick) (we are considering using screenshots or quotes). If you would like to be mentioned in a certain way or don’t want to be mentioned at all, please let us know, we will always follow up on your request.

And of course, if you have any general questions or remarks ...

Thanks for helping us !!
  • Like
Reactions: drklu
like this
Hi Mr. H!

My first question - what kinds of flow rates do you expect this mechanism will net you? We typically aim for 1GPM or around 240LPH in custom loops. That's honestly kindof overkill, but we tend to err on the side of overkill.

Also, compared to water, does your coolant have a comparable specific heat capacity? Is it chemically compatible with copper? Brass? Nickel? PETG plastics? Acrylics? Does it rely on a specific geometry in the loop architecture or could it be put to use in a custom built loop?

Sorry to bombard, but this sounds really interesting!
Hi VanGoghComplex !

Thanks for your reply !
Unfortunately, we don't have a prototype yet, therefore, it's all calculations for now.
The flow in our system increases with the difference in temperature between ambient and the heatsink temperature.
With lets say an ambient temperature of 20 degrees and a heatsink temperature of 70 degrees, we should obtain a flow of approx. 90 LPH.

The speciic heat capacity will be less than that of water, but in the same order of magnitude, so the difference should not be too big to have a huge effect.
With respect to the compatibility, the liquid is oil-based and therefore compatible with all metalics. Furthermore, additives can be added to increase the compatibility.
But when it comes to plastics, this could cause some issues and we definately need to do more research on this part too.

This is what we know so far. Finding out if its something that warrants further research and potential investments is one of our goals here, so we are really interested in your thoughts! Feel free to bombard us any time!
I'm not the foremost authority here, but if the numbers you provided are accurate, then I believe this system would be of limited application to a PC enthusiast. It might certainly have a place, but if our (meaning we enthusiasts) heatsinks are 70c, we are already very unhappy. That means our components are several dozen degrees hotter than that and we are already thermal throttling.

This of course is assuming you didn't obfuscate your numbers above to protect your research (which I would totally understand if you did).
I'd argue that the concern here is the higher Delta T resulting in more effectiveness. I mean this is the case for all heat transfer, from what I recall of high school physics, but unless the heat transfer ramps up rather rapidly with Delta T, I'd be concerned that the steady state temperature under load might be quite high.

Some of our systems generate over 200W at load and it's all dissipated through an itty bitty 75mm^2 chip under those metal heat spreaders, and we want those core temps to stay below 70°C.

It will all depend on the specifics though. If it doesn't get up and going rather quickly with Delta T,it might still be OK for mobile and other low power solutions.

How much of a Delta T do you think this magic fluid of yours needs to get moving effectively?
To answer the op's actual questions:

Performance of a PC cooling system is, realistically, measured in several ways:
1. Actual cooling - that is, how low can you get the temperature of the IC you're trying to cool. Assuming there's no phase change involved, water and glycol are the benchmark.
2. Compactness - The ability to fit an adequate cooling inside the case is a major limiting factor in most systems to some degree, but particularly in smaller systems. The problem is usually whatever heat exchanger dissipates the heat into the air such as a radiator or the end of the heat pipes with the fins, which often has to be quite large.
2.5 - Other practical considerations - You mentioned that the coolant is an oil-based substance. I think this will be a big turnoff for most, especially if the coolant is toxic or flammable.
3. Cost - A good water cooling setup can easily cost $500. CLCs are less expensive, but still in the $100 to $200 range, generally. If you can beat this cost, and also beat air heatsinks in terms of performance, you may have a hit.
4. Aesthetic considerations - If we're being totally honest, some of us do water cooling mainly because it looks cool, although I expect few would be willing to admit it, in much the same way that cyclists like to talk about how shaving their legs makes them faster, but they really just do it for looks.
5. Reliability - Within reason, this is likely to be a low priority to most enthusiasts, but will be a huge deal to enterprise customers for this technology. Speaking for myself, I'll deal the potential for (water) leaks or pump failures if it means I get better cooling.

My instincts tell me that the enthusiast market may not be the best customer for this, if, as Zarathustra[H] alluded to, it requires a large temperature delta between the hot end and cold end of the cooling system. A really good water cooling system can get that delta down to about 20C, which is important because the speed the computer runs at is directly related to that difference in temperature. Most enthusiasts will live with the potential for pump failure if if means we get that as low as possible.
This sounds a lot like thermosyphoning to me. Using density change in cold and hot fluid to promote flow with buoyancy.

As stated above, I think the big concern is the required delta T. If this is what you're up to, I'd recommend looking at the nuclear industry. This strategy is how some reactor designs are able to passively cool themselves. Requires the boilers being higher than the reactor. The idea is to buy some time after a total loss of power. When the main coolant pumps trip, there's a flywheel in them to keep them spinning for a short time to help promote flow.

Not sure how well this is going to work for electronics. If this is the plan, I'm thinking that the required delta T will result in way too high temperatures.

But hey it's your project, I'm definitely interested!
As others have said, flow rates are far too low to be of practical usage to us. It may make sense where large delta Ts are fine, but one of the main points of watercooling is minimizing the delta T.
For computer cooling you need to consider how setup is laid out.
This setup i think is enforcing that hottest components should be on the bottom of the loop.

This kind of setups would likely demand whole "case" to be submerged

immersion cooling?
As I see it there are currently three major types of liquid cooling, listed below, and a new product should bring some sort of improvement to any one of these.

1. Heatpipe.
Fixed format cooling where the heat is transfered a fairly short distance from hotspot to cooling fins. Cheap and efficient.

2. AIO cooler.
Closed system for cooling of one component. Somewhat soft hoses allow some flexibility in the placement of the convector. Need a pump.

3. Custom water cooling.
Typically used to cool more than one component. Highly flexible in layout options. Convector size limited only by cost and available volume. Coloring of the water is common and often part of the reason for using water cooling in the first place. Expensive and require some maintenance.

Some decade ago there was a new cooler on the market that looked like a heatpipe cooler but used some magnetic properties to pump a liquid in a loop inside. It added no significant performance over heatpipe but was more expensive, and didn't survive on the market.
I am a student at the University of Technology in Delft.

While Google is... two clicks away, it might be worth mentioning where you're from, what program you're with (basic background stuff) as well as why you're interested in this particular problem and how you arrived at the [H] forums to ask the question -- as well as where else you've asked. Not asking for you to swear allegiance to some idea or trying to scare you off, just since your account is new, a little background can help with community engagement.

That all said -- from the responses above, it sounds like the flow rates are going to be too low for higher end desktop setups, while at the same time, costs are going to be too high for lower-powered setups. The draw of less noise by subtracting the pump is definitely nice though, and your project could very well be tested as a pump-less closed-loop cooler, just so long as you accept that your ΔT will likely fall short of supporting enthusiast-grade CPUs and GPUs.

At least at first!