cpu CPU cooling idea

[invalid-user]

im interested about whacky cooling ideas like when linus cooled his pc with an ac unite or with a pool.

but that isnt crazy enough for me, i thought about directly cooling the cpu die and im not talking about using those standard off the shelf direct die water blocks.

im talking about CPU die making physical contact with a cooling liquid directly.

 

like i think it would be easy enough to make and test out with an old core 2 duo pc.

 

and my idea is like this (look at image below)

 

and this isnt a new idea it has been done before with special liquids but thats because that liquid is basically a non sticky liquid that drys auickly leaving no residue for frequent check ups.

 

ao tell me why linus cant make this a video

 

because if i got a few 100 bucks lying around id definitely make it.

[TatamiMatt]

9 minutes ago, invalid-user said:

im interested about whacky cooling ideas like when linus cooled his pc with an ac unite or with a pool.

but that isnt crazy enough for me, i thought about directly cooling the cpu die and im not talking about using those standard off the shelf direct die water blocks.

im talking about CPU die making physical contact with a cooling liquid directly.

 

like i think it would be easy enough to make and test out with an old core 2 duo pc.

 

and my idea is like this (look at image below)

 

and this isnt a new idea it has been done before with special liquids but thats because that liquid is basically a non sticky liquid that drys auickly leaving no residue for frequent check ups.

 

ao tell me why linus cant make this a video

 

because if i got a few 100 bucks lying around id definitely make it.

Because, mainly, this will keep the cpu cool, for all of x amount of time until the liquid boils, or becomes hotter than the CPU and turns into an insulator, cpu overheats thermal throttles and the cooler is made redundant. This is essentially, what a water cooler does, except, regardless of whether the water block is direct die or just on the IHS, it moves the liquid around a ton and to a place where is can cool itself off easily

 

Think of a kettle, the water is the cooling liquid and the heater at the bottom is the cpu, the kettle gets much hotter than it does cold until the water boils

[Hinjima]

4 minutes ago, TatamiMatt said:

Because, mainly, this will keep the cpu cool, for all of x amount of time until the liquid boils, or becomes hotter than the CPU and turns into an insulator, cpu overheats thermal throttles and the cooler is made redundant. This is essentially, what a water cooler does, except, regardless of whether the water block is direct die or just on the IHS, it moves the liquid around a ton and to a place where is can cool itself off easily

 

Think of a kettle, the water is the cooling liquid and the heater at the bottom is the cpu, the kettle gets much hotter than it does cold until the water boils

Yeah but what if the water is actually flowing over the die itself just like with a normal liquid loop and not just sitting there.

I still think a cold plate is more efficient due to larger surface area the water will hit.

[RONOTHAN##]

6 minutes ago, invalid-user said:

im talking about CPU die making physical contact with a cooling liquid directly.

This is a bad idea.

 

First off, there's capacitors on the top side of the CPU substrate, so there are solder joints that have voltage across them, therefore using a liquid like water would bridge these contacts and cause a short. You'd need to use a non-conductive liquid, but the water based ones become conductive over a relatively short period of time, and the non-water based ones have horrific thermal transfer properties making their performance terrible. 

 

Second, there's no microfins. One of the reasons that waterblocks are a thing in the first place is that they have a bunch of very tiny channels for the water to go through, increasing the surface area that the waterblock has tenfold, and because higher surface area means higher performance, they perform significantly better than just a flat metal box like the original waterblocks were. The CPU die does not have these microfins, it's just a flat and therefore would suuuuck at transferring its heat to a liquid, you really want to have a waterblock bonded to it and using the microfins for their cooling benefits. 

 

There's probably more reasons, but either one of those two alone is enough to put an end to a project like this. 

[TatamiMatt]

8 minutes ago, Hinjima said:

Yeah but what if the water is actually flowing over the die itself just like with a normal liquid loop and not just sitting there.

I still think a cold plate is more efficient due to larger surface area the water will hit.

If its actually flowing, and im not sure how that would work with dead spots etc, and would be a semi complex bit of fluid mechanics and positioning on the board etc. Ignoring that, having all 5 exposed sides of the silicon die used as a surface to extract heat from would be some pretty interesting cooling (not great, but certainly interesting), like for example a direct die straight to liquid cooling block

 

But as @RONOTHAN## said, I have many doubts about this, though it is an interesting thought experiment, you dont make progress without trying new and maybe whacky stuff

 

Said whacky stuff/quick design below just for funsies

 

red: cpu

blue:tubing+block

aqua:water/coolant

solid line:flow around

dashed line: flow over top

 

[jaslion]

It's been done with mineral oil submerged pc's. Even with a JET flow over the cpu it couldn't get enough cool liquid in time and just overheated to shit.

 

Waterblocks have a VERY important task and that is spreading the heat out over as much surface area as possible as fast as they can. A cpu ihs is NOT good enough for that and won't suffice even remotely close.

 

 

[Thaldor]

It's called immersion cooling and pretty well researched thing. The main problem with it is the liquid:

Water: If not 100% pure conducts electricity and things die. Also quite bad at thermal conductivity so needs more surface area than the IHS of CPU can provide (this is why waterblocks have microfins and you have the pretty dense cooler).

 

Mineral oil: The good old immersion cooling liquid. Doesn't conduct electricity so you can dump your whole PC in it and it will work. Even worse at thermal conductivity than water so you need fans to push that oil through air cooler to keep moving the heat away from the CPU.

 

Something like car engine coolant: It's basicly water with some chemicals to make it so it doesn't corrode, freeze and nothing can live in it (so you don't get algae in your cooling loop). Basicly PC water cooling additives turn the distilled water mild engine coolant.

 

Exotics like M3 Novec: Expensive and gets really engineery really fast. M3 Novec while having good thermal conductivity and boiling point suitable for direct immersion cooling (as shown by Der8auer) has the problem of being too liquid and too gassy, you need special cables and/or connectors for the liquid not flooding your cables and you need to make the PC pretty air tight so not a lot of the liquid escapes after being turned into gas bu the heat. Generally these also have the problem of the boiling point, too high and they won't transfer heat fast enough, too low and as the CPU heats up there starts to be Leidenfrost effect where liquid cannot get to the CPU to cool it (the liquid boils so fast it creates gas layer between the IHS and the liquid).

 

There also isn't magic liquid that would just make the heat go away. Some old mineral oil PC's could work without additional cooling but for anything modern and more power hungry, you are going to need secondary loop that will cool the immersion liquid. So, either condensing loop as Der8auer with M3 Novec where there's cooler that runs cold water (from another cooler that is air cooled) on which the M3 Novec condenses again and drips down to be heated again, or just cooler immersed into the liquid that cools it down.

 

Only real thing you can do here is to have better idle reserve. As in, if your CPU is immersed into 10L of mineral oil, it will take a long time for that oil to reach the 90C where it won't cool the CPU anymore or it wil ltake a long time for 3L of M3 Novec to boil to the point where there isn't enough liquid to cool your PC. Or at least those will take much longer than it takes for your CPU waterblock with like 10ml of water and the copper to heat up to the 90C and your CPU shutsdown or the aircooler to completely saturate and same thing happens.

[invalid-user]

2 hours ago, TatamiMatt said:

Because, mainly, this will keep the cpu cool, for all of x amount of time until the liquid boils, or becomes hotter than the CPU and turns into an insulator, cpu overheats thermal throttles and the cooler is made redundant. This is essentially, what a water cooler does, except, regardless of whether the water block is direct die or just on the IHS, it moves the liquid around a ton and to a place where is can cool itself off easily

 

Think of a kettle, the water is the cooling liquid and the heater at the bottom is the cpu, the kettle gets much hotter than it does cold until the water boils

ik

obviously with a radiator a pump and a reservoir 

no5 just a cup of water

[Hinjima]

19 minutes ago, invalid-user said:

ik

obviously with a radiator a pump and a reservoir 

no5 just a cup of water

 

[invalid-user]

5 hours ago, Hinjima said:

 

thats not what i had in mind

im talking about the actual cpu die making physical contact with the cooling liquid 

be it oil or demineralized water....

[invalid-user]

8 hours ago, Hinjima said:

Yeah but what if the water is actually flowing over the die itself just like with a normal liquid loop and not just sitting there.

I still think a cold plate is more efficient due to larger surface area the water will hit.

bro youre the first one to understand what im saying XD

 

like obviously with a water mup reservoir and a radiator to cool the liquid

 

like if the heat has to go from the heat spreader through thermal past through the water block and then to the water that would be imo a lot of waste

 

why not just have the liquid touch the die directly and yes maybe add a few micro fins... but is that really necessary?

 

 

[Hinjima]

41 minutes ago, invalid-user said:

bro youre the first one to understand what im saying XD

 

like obviously with a water mup reservoir and a radiator to cool the liquid

 

like if the heat has to go from the heat spreader through thermal past through the water block and then to the water that would be imo a lot of waste

 

why not just have the liquid touch the die directly and yes maybe add a few micro fins... but is that really necessary?

 

 

Hehe, yeah I know exactly what you are thinking

Its good in theory but in reality I do believe its a bit more complicated.

Adding fins or grooves to the Die itself would increase the cooling potential dramatically but this has to be done from factory by the Die producer.

[Thaldor]

20 minutes ago, invalid-user said:

why not just have the liquid touch the die directly and yes maybe add a few micro fins... but is that really necessary?

Let me put it this way:

Air has thermal conductivity of 0.03 W/mK, water has thermal conductivity of 0.61 W/mk and copper has 384 W/mK (higher the number the better).

Thermal conductivity tells you how much thermal energy moves through a material and at the same time how well the material removes thermal energy from around it. However why water cooling is better than air cooling is heat capacity which means how much thermal energy the material can hold and while copper and any metal cleans the floor with water and pretty much any liquid (excluding liquid metals), water is pretty good at holding that heat.

Air and water can move as much heat energy as copper but to do so they need a lot more surface area to suck it in.

(there's also materials that have good thermal conductivity and heat capacity but they have other quite huge problems, like the liquid metals which are poisonous or oils which are viscous and messy.)

 

This is why you use metal heatsink that can move a lot of heat from the rather small surface area of the CPU die and transfer it to the fins which have a ton more surface area so water or air can dissipate that heat into themselves (water to move it to the cooler that has even greater surface area to transfer that heat to air, which you usually have a lot so your CPU must release a lot of heat to heat the room full of air even 1C).

 

This is also why phase change cooling is so powerful because instead of using some intermediate material to move the heat from a liquid to gas, it just lowers the pressure to change the liquid into gas and still keep all the heat energy in it and then it's cooled and pressurized again back to liquid which can be much colder than the gas and has much better thermal conductivity. Phase change is also why something like 3M Novec works because it's boiling point is around 70C, it phase changes from liquid to gas in 70C so it can directly in normal atmospheric pressure (because pressure affects the boiling point of a liquid) work as phase change cooling liquid.

(Too bad 3M Novec has so many other problems like being too thin and actually managing to get in and flood stuff like HDMI cables, being expensive, also because very thin needing a ton of insulation, like your normal water tight is not even close to be enough, and because of that it would have been even more expensive. Also not all plastics and whatever were really happy about 3M Novec because it's a organic compound, also it's possibly quite poisonous in certain forms.)