Liquid metal - extreme cooling

[LED_Guy]

Has anyone ever tried liquid metal cooling? You loose some efficiency since metals have a lower heat capacity, but you make up for it in spades with thermal conductivities that are >100 times higher.

You can also use an electromagnetic pump (silent because the only moving part is the liquid metal). If you use metal lines then these act as radiators too - although the alloy I am thinking of using makes an amazingly good mirror.

Liquid metal cooling is used for extreme cooling (like nuclear reactors). There are alloys with melting points well below room temperature (that don't require mercury).

I'm thinking about trying a to build a small CPU cooler based on the concept. Thoughts?

[WoodenMarker]

That sounds cool--literally.

Reading the title, I thought you were referring to coollaboratory liquid pro/ultra.

 

I used CLP for my delidded 3770k but nothing like what you mentioned.

[Guest]

That sounds cool--literally.

[Fate]

Woah, sounds really cool.

[LED_Guy]

I could build a pretty decent gaming rig for what it will probably cost me for a system that will handle the CPU, GPUs and VRMs, but it would be hands down the quietest system around.

One of the potential advantages is that you don't need nearly the same amount of fin surface area. I might be able to use the case panel behind the mobo as the "external" radiator.

[LED_Guy]

Do better searches. Gallistan and indalloy are commercially available alloys. I worked with indium/gallium alloys for electrical contacts in the past (research project for Ph.D. In Chemical Engineering). I'm not too worried about the alloy. I have some reasons to slightly modify it.

I need to do some calculations to determine exactly how much more efficient it will be compared to water. That lets me estimate the flow rate, radiator design and total liquid alloy volume.

I expect that I might need to add brackets to stabilize my GPUs if I get that far on the project.

[LED_Guy]

With stainless steel, gallium isn't too much of a problem at temperatures below 300 C. If I'm in that neighborhood I have a lot more problems. Standard rule of thumb is reaction rates change by 2x for every 10 C temperature change. That means corrosion would be 2^-20 less of a problem (that's about a factor of a million below what is considered a concern).

Since I already said I would be using an electromagnetic pumps, why do you say that corrosion there would be a problem? An electromagnetic pump needs two electrodes, a magnetic field and a current. The body of the pump can even be glass. The only thing that moves is the liquid.

A little more research before shooting down the idea would be appreciated.

[TheIcedCanadian]

If you make this make sure to make a build log and tag me - looks like a sick awesome idea and I can't wait to see it!

[LED_Guy]

Will do. Testing the pump concept will be the easy part. I have enough indium and gallium lying around to make a few hundred cc's of coolant + an hour or two of calculations to tell me I'm not crazy.

After that it will take about 6 months to design and build the rest of the system. If I go forward I will definitely post the build. I'll probably end up spending at least $2,000 on it, but it might be fun . . .

I came up with the idea after watching Linus's silent PC build video . . .

[d33g33]

I used CLU after delidding my 3770k Thats as far as im going

[LED_Guy]

Well, if you're willing to de-lid, then you should look into a passivation coating (silicon dioxide or more preferably silicon nitride).  That will protect the circuitry.  Run liquid cooling directly on that and you will should see some major improvements.  That takes out most of the resistances.

[Ghost]

Liquid metal cooling is used for extreme cooling (like nuclear reactors). There are alloys with melting points well below room temperature (that don't require mercury).

Liquid metals in nuclear reactors are used because they can cool very hot things much more easily. I'm going to cheat here and quote from the wiki:

 

"They have safety advantages because the reactor doesn't need to be kept under pressure, and they allow a much higher power density than traditional coolants. Disadvantages include difficulties associated with inspection and repair of a reactor immersed in opaque molten metal, and depending on the choice of metal, corrosion and/or production of radioactive activation products may be an issue."

 

And why water is not used:

 

"While pressurised water could theoretically be used for a fast reactor, it tends to slow down neutrons and absorb them. This limits the amount of water that can be allowed to flow through the reactor core, and since fast reactors have a high power density most designs use molten metals instead. Water's boiling point is also much lower than most metals demanding that the cooling system be kept at high pressure to effectively cool the core."

 

As you don't face any of these issues I don't see why you would not just use water.

 

It seems like a lot more effort than phase change cooling except with phase change you will be able to go below room temperatures.

[LED_Guy]

Standard liquid cooling uses water which has a high heat capacity, but a low thermal conductivity. This means it takes a lot of heat for a small temperature change. In contrast liquid metals have a lower heat capacity, but tremendously high thermal conductivity. It doesn't take a lot of heat to increase the temperature, but the heat spreads through the liquid metal very quickly. Think of it as a liquid heat sink.

With a liquid metal the heat can spread faster than the liquid flows. When a liquid metal heats up, all of it heats up, not just the part next to the component (that's what happens with water).

The whole point is better cooling and no noise. The efficiency can be a lot higher than even phase change. Higher efficiency also means a smaller (or no) radiator. Would you turn down a liquid metal cooling system if it could handle your CPU, multiple GPUs and VRMs with a single 120mm fan and radiator set? What if the liquid metal was run along the inside of the case and used the outside of the case for a radiator and no fan at all? Talk about a clean and quiet build! Free up space in your case and get rid of noise.

At least to me there is a real appeal to a pump with no moving parts (no noise, nothing to fail). The only thing that moves will be the liquid metal.

You seem to be in favor of using fans over liquid cooling. Why not just use a 386 CPU and not worry about cooling at all?

Oh yeah, then there is the fact that it would just be a really cool (no pun intended) thing to do.

[Ghost]

Standard liquid cooling uses water which has a high heat capacity, but a low thermal conductivity. This means it takes a lot of heat for a small temperature change. In contrast liquid metals have a lower heat capacity, but tremendously high thermal conductivity. It doesn't take a lot of heat to increase the temperature, but the heat spreads through the liquid metal very quickly. Think of it as a liquid heat sink.

With a liquid metal the heat can spread faster than the liquid flows. When a liquid metal heats up, all of it heats up, not just the part next to the component (that's what happens with water).

The whole point is better cooling and no noise. The efficiency can be a lot higher than even phase change. Higher efficiency also means a smaller (or no) radiator. Would you turn down a liquid metal cooling system if it could handle your CPU, multiple GPUs and VRMs with a single 120mm fan and radiator set? What if the liquid metal was run along the inside of the case and used the outside of the case for a radiator and no fan at all? Talk about a clean and quiet build! Free up space in your case and get rid of noise.

At least to me there is a real appeal to a pump with no moving parts (no noise, nothing to fail). The only thing that moves will be the liquid metal.

You seem to be in favor of using fans over liquid cooling. Why not just use a 386 CPU and not worry about cooling at all?

Oh yeah, then there is the fact that it would just be a really cool (no pun intended) thing to do.

As metals have a lower thermal capacity you actually need to move the coolant through the block much faster -> louder pump.

 

Also, waterblocks are designed to give as much surface area inside the block touching the water, so water isn't that bad of a coolant inside them and also cause turbulence in the flow. (They push the water through very narrow slits.)

 

Using an electromagnetic pump is interesting but I would worry about inducing current in the coolant, which would be transferred directly to the die of your CPU/GPU.

 

I have no idea where I said I am in favour of fans, although liquid cooling requires fans anyway so I guess I am?

 

An yes it would be "cool" but it's not a practical form of cooling, there are easier, tried an tested, more effective solutions. If you really don't like water as a coolant you can use direct die phase change cooling (the evaporator touches the die directly.)

[Skander1345]

Go for it. I'd like to see the results for myself instead of theories.

[TheIcedCanadian]

I've got no idea WTF is going on, but I like it

[SurvivorNVL]

If you actually manage to build it; try and gain a companies attention and then wave your arms about and say, "Buy this and market it!"  I love this idea.  I personally will keep on air cooling until Phase-change becomes more affordable like what Captherm Systems is trying to do with their MP1120.

[LED_Guy]

There is a balance with water cooling. It may have a high heat capacity, but the low thermal conductivity means that you only heat the "skin" of water next to the cooling block. You get around this by designing for turbulent flow. That means high flow rates through narrow channels. You don't necessarily need turbulent flow for liquid metals, so you might have a lower flow rate. Keep in mind that metals also have a higher density, so volumetric basis for heat transfer is better than simple heat capacities would suggest.

As for electric currents, if the electrodes are in the "pump" section, how are they going to get to my CPU or GPU? To begin with you would have to get the voltage inside the package. The packages are designed to prevent that. Even without that the current would flow between the electrodes in the pump section and no go anywhere else unless you majorly screw up the grounding in the system.

Did you miss the Ph.D. In Chemical Engineering part? I have a lot better technical background in in heat transfer than most people.

I appreciate the tips if that's what they are, but I am starting to get the feeling you think I am wasting my time and that it is a less than intelligent way to spend my time and money. It is my time and money though.

I guess that no one would be interested in a cooling system that would cool all your components without cluttering up a case and requiring huge radiators everywhere. That's what I am trying for: a high capacity cooling system that doesn't require lots of radiators, will be absolutely silent and won't add lots of hardware to your case.

Think about a 5960X system with 2-3 GPUs in SLI/Crossfire with everything over clocked all in a mini tower.

 

Oh yeah, there is the whole push the envelope thing. Horses and carriages were "tried and true" once upon a time. So we're 8086 processors, monochrome screens and magnetic tape.

Who needs multi-core processors, SLI/Crossfire GPUs, SSDs and 4k monitors?

 

Thanks. Marketing isn't really my goal. If this works then it will be an expensive option that most users won't be able to afford. The will be some who will still want to do it, just because it's cool (no pun intended) or because it is the highest performance cooling system possible.

This is a forum for enthusiasts, so I am sharing what I am trying. It's a lot more fun for me than working in isolation.

The are some technical points/issues that haven't been brought up that I am already aware of and I have some potential solutions. Still I might need some help . . .

[B NEGATIVE]

There is a balance with water cooling. It may have a high heat capacity, but the low thermal conductivity means that you only heat the "skin" of water next to the cooling block. You get around this by designing for turbulent flow. That means high flow rates through narrow channels. You don't necessarily need turbulent flow for liquid metals, so you might have a lower flow rate. Keep in mind that metals also have a higher density, so volumetric basis for heat transfer is better than simple heat capacities would suggest.

As for electric currents, if the electrodes are in the "pump" section, how are they going to get to my CPU or GPU? To begin with you would have to get the voltage inside the package. The packages are designed to prevent that. Even without that the current would flow between the electrodes in the pump section and no go anywhere else unless you majorly screw up the grounding in the system.

Did you miss the Ph.D. In Chemical Engineering part? I have a lot better technical background in in heat transfer than most people.

I appreciate the tips if that's what they are, but I am starting to get the feeling you think I am wasting my time and that it is a less than intelligent way to spend my time and money. It is my time and money though.

I guess that no one would be interested in a cooling system that would cool all your components without cluttering up a case and requiring huge radiators everywhere. That's what I am trying for: a high capacity cooling system that doesn't require lots of radiators, will be absolutely silent and won't add lots of hardware to your case.

Think about a 5960X system with 2-3 GPUs in SLI/Crossfire with everything over clocked all in a mini tower.

 

 

Already been done,you can run all that on a 280 rad with 1000 rpm fans,dont get suckered into the excessive rad mentality on this forum.

 

Your idea has merit but is overly complicated and has inherent H+S issues. Water is cheap,simple and effective,if thats not enough then TEC/Phase are much more effective. You are still bound by ambient temps even with the wunderkind liquid metal,the coolant is not the issue,the reduced die sizes of current Intel chips lead to hotspots on the chip,heat density in a small area is the problem now,not the coolants ability to move the heat.

 

 

I have to ask,if you have these diploma's,why you are asking for help here?

[LED_Guy]

To begin with:  I thought the concept might be interesting to some of the readers of the forum.

 

Secondly: despite the diplomas, I have never been one to believe that I have all the answers.  Too many engineers fall into that trap, and at times it makes for some spectacular failures.

 

As you point out, the smaller die sizes of Intel (and AMD) chips presents more of a problem.  The issue comes down to a power density (W/mm^2).  As the power density increases, it actually becomes easier to remove heat (due to larger temperature gradients), but the whole issue is keeping the die temperature low which means a lower temperature gradient.  There's no free lunch here.  If you can't increase the chip thermal limit/operating temperature you have to go to something better.

 

You're statement about hot spots on the chip is directly related to the coolant.  Water has a low thermal conductivity so you have to have high flow rates to move the heat.  Even then the heat only moves in the direction of the flow (basically one dimensional cooling).  With a liquid metal you can move the heat much more efficiently in all directions.  You have a mobile heat sink.

 

BTW: gallium and indium are not really H&S concerns.  I suppose they might be if you swallow them, but even then they would be much less hazardous than lithium batteries.

[B NEGATIVE]

To begin with:  I thought the concept might be interesting to some of the readers of the forum.

 

Secondly: despite the diplomas, I have never been one to believe that I have all the answers.  Too many engineers fall into that trap, and at times it makes for some spectacular failures.

 

As you point out, the smaller die sizes of Intel (and AMD) chips presents more of a problem.  The issue comes down to a power density (W/mm^2).  As the power density increases, it actually becomes easier to remove heat (due to larger temperature gradients), but the whole issue is keeping the die temperature low which means a lower temperature gradient.  There's no free lunch here.  If you can't increase the chip thermal limit/operating temperature you have to go to something better.

 

You're statement about hot spots on the chip is directly related to the coolant.  Water has a low thermal conductivity so you have to have high flow rates to move the heat.  Even then the heat only moves in the direction of the flow (basically one dimensional cooling).  With a liquid metal you can move the heat much more efficiently in all directions.  You have a mobile heat sink.

 

BTW: gallium and indium are not really H&S concerns.  I suppose they might be if you swallow them, but even then they would be much less hazardous than lithium batteries.

 

 

You are barking up the wrong tree,coolant is not the problem,most simple loops run within a 5-8c delta with water. Your liquid metal may shave a few c off,thats all. The issues regarding small dies and cooling is directly attributed to the small surface area and density of the die,not the coolant.

 

Galluim and Indium metalloids are known to have several toxicities and to cause carcinogenesis in animals and humans,not something for home use.

Galinstan would be a better choice,speak to Rockwell Collins about a sample.

And 1 GPM is nothing....why are you obsessed about flow rates?

 

 

 

While I am interested in your proposal,I cant help but think you are trying to re-invent the wheel.....for no real benefit.

You dont have an answer for the ambient nature of the cooling either.........

 

 

I look forward to a PoC from you soon.

[Leonard]

Yes this does sound very very cool but at what cost can i, you, achieve this. my piggy bank is over flowing but still.....

[LED_Guy]

Galistan is an alloy of gallium, indium and tin.  Saying that galistan is safe and indium and gallium are not makes no sense what so ever.  So gallium and indium are hazardous by themselves, but not mixed together?  I have read through dozens of MSDSs on both materials, neither is carcinogenic or mutagenic, although if you check the MSDS's for them there are comments about intravenous exposure. . .  Injecting liquid metal into your blood stream takes a very deliberate effort and is more of a mental issue than a safety issue.

 

The whole point I have been trying to make about flow rates is that liquid metal doesn't need anywhere near the same flow rates as water cooling.  Flow rates gets to Reynolds number and turbulent vs laminar flow, mass transfer/thermal boundary layer thicknesses and thermal resistance.  It's all tied up in general transport phenomena.

 

With water cooling only the water directly above a hot spot is able to remove heat due to the low thermal conductivity of water.  Only a thing layer of water directly above the hot spot actually absorbs any heat.  Liquid metal means that the heat spreads much faster so a much larger volume of material participates in the cooling (thick layer that extends past the boundaries of the hot spot).

 

Some times even a few degrees C can make a difference.  I will note that I see a lot of comments about a few extra MHz or a few extra fps in a game.  What do you think running a CPU or GPU a few C cooler translates too?

[B NEGATIVE]

Galistan is an alloy of gallium, indium and tin.  Saying that galistan is safe and indium and gallium are not makes no sense what so ever.  So gallium and indium are hazardous by themselves, but not mixed together?  I have read through dozens of MSDSs on both materials, neither is carcinogenic or mutagenic, although if you check the MSDS's for them there are comments about intravenous exposure. . .  Injecting liquid metal into your blood stream takes a very deliberate effort and is more of a mental issue than a safety issue.

 

The whole point I have been trying to make about flow rates is that liquid metal doesn't need anywhere near the same flow rates as water cooling.  Flow rates gets to Reynolds number and turbulent vs laminar flow, mass transfer/thermal boundary layer thicknesses and thermal resistance.  It's all tied up in general transport phenomena.

 

With water cooling only the water directly above a hot spot is able to remove heat due to the low thermal conductivity of water.  Only a thing layer of water directly above the hot spot actually absorbs any heat.  Liquid metal means that the heat spreads much faster so a much larger volume of material participates in the cooling (thick layer that extends past the boundaries of the hot spot).

 

Some times even a few degrees C can make a difference.  I will note that I see a lot of comments about a few extra MHz or a few extra fps in a game.  What do you think running a CPU or GPU a few C cooler translates too?

 

 

This is not how it works,the water cools the block,not the chip,heat radiates in to the block structure which is removed by the water,hence the much larger than the die water path. Having used direct die coolers before and seen worse performance (DT Sniper Direct block) than the equivalent copper block,Im not sure you understand the mechanics behind it

 

Running a CPU a few c cooler does .......absolutely nothing.

Especially when your modern CPU is running 30c @ load cooler than a stock cooler with watercooling alone.

 

And again,im not sure why you reference flow rates when they are irrelevant .......

[SuperPug]

I'm thinking about trying a to build a small CPU cooler based on the concept. Thoughts?

 

Which liquid metal did you have in mind? It would need a melting point thats below room temperature otherwise the whole 'loop' thing might not work too well. That kinda leaves Gallistan and Sodium-Potassium Alloy. Galistan costs 200$ for 50 grams and Sodium-Potassium alloy is extremely reactive with air and water so that might not be the greatest idea.

Gallium would be the next best candidate but has a melting point of around 30 Celsius. 300$ will get you 500g of that stuff but i still think its impractical and probably wont work since you might not even be able to get all the gallium in the loop to liquify in time to cool the components before overheating.

 

Dont think its gonna work... or am i missing something, im not a chemist or anything

 

Edit: didnt read the part about you being a chemical engineer so i guess all my points are mute heh