Jump to content
Sign in to follow this  

Two phase immersion cooling

Recommended Posts

After building my first custom loop water cooled pc I would like to build my next rig with immersion cooling.


I'm not interested in oil based solutions but in 3M novec. There's not much sources about this and even less prefabricated parts out there. 

3M has published a nice technical paper on the basics. https://multimedia.3m.com/mws/media/1010266O/3m-two-phase-immersion-cooling-best-practices-technical-paper.pdf

I have found couple local firms that do custom aquariums so that's something I need to look into. If I can find someone who can create the case for me I should be able to do the rest of it myself.

The case would be something like this made from steel frame and glass panels.


The end result would look something like this:

I'm still content with my current build so I won't be actually building this until few years.

Has anyone any experience on the matter or found any good sources about this?

Edited by Drakhorn

Share this post

Link to post
Share on other sites

I know they had a setup made with mineral oil but that's not what I'm interested in. Couldn't find any novec setups at least with a quick search.

Share this post

Link to post
Share on other sites

I think it's time to start thinking more about this and I think I will be using this thread to collect intel and set up a plan. Most of the info here will be very custom for this individual build but maybe it will help out others who might be interested in this.

I still haven't decided if I will execute this build or not. I guess it comes down to price and difficulty of the build.

Der8auer has couple of good videos on the subject and they gave me couple of ideas that I think will implement as well.


The case

I have found a firm that crafts custom aquariums: https://www.akvaariokauppa.fi/new/p7947-mittatilausakvaariot-fi.html

- They said they only build aquarium using acetoxy silicon sealant, I have sent a question to 3M about compatibility of that stuff

Other option would be to order a one piece glass container from a glass blowing shop: https://lasi-kostamo.fi/

One piece container would remove the trouble with edge sealing...

- They estimated that a one piece glass container of my required size could cost even as much as 800€ depending on the molds their subcontractors have available

It is becoming clear that any os the adhesive sealants used in aquarium making are not adviseable to use with Novec 7000. Options are to order a custom made one piece glass container or craft the tank using a metal frame and attaching the glass panels with bolts and using gaskets as a sealant.


To keep the case simple I was thinking of ordering a regular glass aquarium with custom dimensions and craft all the hardware mounting points as well as the cable and tube passthroughs into the lid. That way lifting the components from the case should be doable with lifting the lid off the aquarium. I was also thinking of copying Der8auer and adding an exhaust port into the case with another cooling element in it. That way I shouldn't have to worry about overpressure and still keep the gasses inside the case when the cooling element liquifies the gasses before they escape.


Freeboard of at least 10cm is recommended. Higher freeboard is better but this must be weighed against power density, system height, accessibility, etc.


...water will migrate to the top of the system into the headspace above the vapor zone. Its concentration in the headspace can rise above the saturation concentration and water will begin to liquefy. Liquid water can cause short circuits and corrosion. A desiccant such as silica gel in a perforated container placed in the headspace will prevent liquefaction of water. ... The desiccant mass should be at least 5 times the anticipated water mass. A system should be quite dry after the first day of operation. As long as it remains operating, the descant is no longer needed but should be replaced upon start-up whenever the system is allowed to cool.


Temperature sensors in the vapor zone can be used to determine vapor height as it may be useful for regulating water flow to minimize movement of the vapor zone. A temperature sensor above the condenser will warm to the fluid boiling temperature if the condenser capacity should be exceeded.



I'm done with the first sketch. All the dimensions are completely made up, this was mostly done just to learn CAD. The final design will have to be done completely from scratch. 


Reference schematics for the backplate:





2nd iteration of the mounting tray:


Overall dimensions of the iteration: 254,94mm x 127,22mm x 311,22mm (not including protrusions of cables and connectors)

In this configuration the case would eat about 11 liter of the cooling liquid which is a hefty price...

Reference guide for cutting tempered glass:


  • All work has to be done before tempering.
  • Diameter of the perforation has to be at least 5mm or with thicker glass at least as wide as the thickness of the glass
  • Diameter of the perforation cannot be more than 1/3 of the width of narrowest side
  • Distance from the edge has to be at least 1,5 times the thickness of the glass for perforations less than 50mm and glass thickness less than 8mm
  • Distance from the corner of the glass has to be 4 times the thickness of the glass for perforations less than 50mm
  • Edges of the perforations has to be rounded


2nd iteration of the case. With steel frame and tempered glass panel that is bolted into the frame and a rubber gasket. I'm not completely satisfied with it, it is not 100% symmetric but gets the point across. Next iteration I should switch the positions of the PSU and storage media on the backside of the mounting plate, measure the dimensions of the SSD trays and design a block on the bottom of the case behind the tray. That should displace about 3 liters of the cooling liquid and save about 650€. Also need to design the lip of the case where the lid can be bolted on to.


3rd iteration of the case. Link to the project: https://a360.co/3q9AA4t

Still missing all the perforations from the lid, handles to lift the lid and a mounting tray on the outside for cooling system. Final tuning still has to be done with correct component measurements


The exhaust condenser could be something like this with a shroud added on top of it to restrict airway out of the case. https://www.alibaba.com/product-detail/Mini-Air-cooled-Condenser-For-Showcase_60657912569.html?spm=a2700.gallery_search_cps.normalList.5.6dc75c15K7ofqw

But wouldn't condensing the exhaust gas also mean it would condense water from the air into the case? -> yes it would and it would still leak some gasses. Better to go with sealed case and bellows for the overpressure.

Good video on fluid loss mitigation solutions: 


Instead of fabricating them myself, I could go with DIY tray parts like these: 




Most low profile GPU mount found so far: https://www.jimms.fi/fi/Product/Show/160509/sc-vgpu-mt/singularity-computers-spectre-2_0-2nd-vertical-gpu-mount-naytonohjaimen-vertikaalinen-kiinnityssarja-musta 

Considering I will be stripping the GPU of its stock cooler, the card will be significantly lighter thus not needing that much of support so maybe go with a very simple solution that is also PCIE 4.0: https://linkup.one/linkup-ultra-pcie-4-0-x16-riser-cable-rtx3080-x570-b550-rx5700xt-tested-extreme-high-speed-itx-vertical-gaming-pci-express-gen4-2020-universal-90-degree-socket-20-cm-3-0-gen3-compatible/

Power switch: https://www.aliexpress.com/item/32676328352.html

Dont forget rubber feet for the bottom: https://www.elfadistrelec.fi/fi/rubber-feet-square-nelioe-20-5x20-5x7-6mm-70-shore-musta-3m-sj-5023-black/p/14886594?q=&pos=8&origPos=8&origPageSize=50&track=true


The liquid

I have found a retailer that told me they sell the liquid to private buyers: https://www.yeint.fi/elektroniikka/kemikaalit/jaahdytys/3m-novec-hfe-7000a-lammonsiirtonest

So the liquid is over 200€ per liter, might wanna look for other sources as well. Better also plan the case efficiently and try to keep it small I guess. I was planning to go with the 7100 liquid with 61C boiling point first as it's easy to work with. While that works with the hottest components it also means that every component in the system will be running at 61C so I'm thinking of going with 7000 liquid now as it has a boiling point of around 35C. Downside is that this requires a constant 24/7 cooling of the exhaust port because the already highly volatile liquid will start evaporating in room temperature quickly and also cooling the gas with room temperature elements might not be sufficient enough to liquify it back into the case, requiring a cooler like peltier plates or a water chiller. Might have to look into an overpressure valve that shuts off the case when the pc is not in use.


Optimal (condenser) designs ... will achieve 4+ kW/liter of core volume at 15°C approach temperature difference (fluid to water) and 10°C water rise.

Good to know when designing the cooling system.


Compatible materiel

Compatible metals:

  • Stainless steel
  • Brass
  • Copper
  • Aluminium

Compatible plastics:

  • Polypropylene (PP)
  • Polyethylene (PE)
  • Nylon
  • Polyacetyl (sic) (POM)
  • Polyether ether ketone (PEEK)
  • Polytetrafluoroethylene (PTFE) aka teflon

Source: https://multimedia.3m.com/mws/media/121372O/3m-novec-7000-engineered-fluid-tds.pdf


(perforations) can be sealed with compression fittings, or minimal use of clean elastomers like butyl rubber or clean pipe thread sealants such as industrial grade PTFE tape or Leak Lock® Blue Joint Sealing Compound

Possible product to be used as a gasket: https://www.thomann.de/fi/monacor_mdm_25.htm?sid=9b96f6d3a5f242c704848b0e7d9c0684


...tanks should not be fabricated from plastics as they contain and transfer moisture that becomes a burden on the system desiccant. Tanks can be fabricated from glass but edge sealing is questionable. Ideally tanks are made from welded metal. Ordinary carbon steel (14 gage for smaller tanks, 11 gage for larger ones) has proven effective, but aluminum and stainless can also be used. Metal should be reinforced as needed and powder coated on the interior and sealing surfaces at a minimum to help prevent oxidation from accidental moisture condensation and to provide a margin of safety against short circuiting to the tank walls.

14 gauge steel is 1.897mm thick.


80 durometer EPDM or butyl should be used with hydrofluoroether fluids (3M™ Novec™ 7000, 7100 or 7200 Engineered Fluids) .... These elastomers contain oils that are inevitably extracted in normal use and are a source of fluid contamination. Elastomers should therefore be chosen that are low in extractable content. Pre-cleaning by vapor degreaser may be indicated to “preshrink” the elastomer and lessen the organic burden.


Most cabling is clad with polyvinyl chloride (PVC) that contains 3-30% oil by weight. This oil is inevitably extracted from the PVC in normal use and is a source of fluid contamination. ... Cabling can be pre-cleaned by extended period (24 hr) vapor degreasing to remove most of the oil, but some will remain and must be removed by the fluid conditioning system to keep the fluid healthy. Cabling clad with polyolefin or PTFE is typically quite clean but more difficult and expensive to acquire.

Heat shrink tubing to shield the cables inside the case:




Cablemods support told me that their clear SFF cables are insulated using teflon. They would be perfect and wouldn't require protective sleeving!

An ultrasonic cleaner could probably be used as a degreased. This I will not include in the budget as missus has been talking about cleaning some jewelry anyway.




If not removed, the hydrocarbon oils found in elastomers, PVC insulation, foams, adhesives, etc. will be solvated by the fluid at low levels and separated by distillation to foul boiling surfaces. It can also diffuse into other materials and cause them to fail. Acid washed activated carbon, similar to the type used in drinking water cartridges, has a high affinity for these materials. Smaller systems can often be cleansed passively by placing carbon filter “bags” in the liquid. ... The amount of carbon used should be at least 5× the mass of anticipated contamination; this can be estimated by totaling the elastomeric mass and assuming that 0.1% (if cleaned) to 3% (typical) or even 25% (extreme) of that mass is contamination. Carbon should be replaced after the first week of service and does not need to be replaced subsequently.



PC Components (as of now, subject to change with upcoming releases)


As a rule, any device that does not have a heat-sink in air requires no modification when immersion cooled. Examples include memory DIMMs, capacitors, etc. Heat sinks should be removed from all other components.


In practice, any device that requires a copper heat sink to be air-cooled effectively will generally require some kind of boiling enhancement coating to facilitate incipience and increase efficiency. Either organic or metallic boiling enhancement coatings can be used. In the case of most lidded devices, such as CPUs in the 50-120W range, the application of a thin porous metallic boiling enhancement coating (BEC) to the lid by soldering or adhesive is suggested. Thermal grease as well as any packages that use it should not be used as the grease can deteriorate after months of service as the oil carrier is extracted. Bare die devices that have heat sinks or lidded ones that exceed 50-120W will require an optimized heat spreader with the BEC applied to it.

The BEC is completely developmental substance and not available commercially. In order not to "choke" the high heat components such as CPU and GPU it is recommended to enlarge the boiling surface area. Pin fin heatsinks made of copper would be ideal but their mounting requires some intuition. Other option would be to attach a water block and take the cap off from them and expose the innards but I think the typical microfins found in them have too small gaps for the liquid to reach the boiling surface. Then there is the issue of TIM. Compounds are out of question I think. I would either have to solder the heatsink onto the IHS or use a thermal pad.

Pin fin heatsinks https://shenzhenlike.en.alibaba.com/product/60731280193-806107620/New_Product_Copper_Pin_Fin_Heat_Sink_Custom_Copper_Pin_Fin_Heatsink_Copper_Copper_Fin_Heatsink.html


Thermal pad: https://fi.rsdelivers.com/product/panasonic/eyga091203m/thermal-interface-sheet-graphite-1600w-mk-115-x/7124009 or https://www.ebay.co.uk/itm/254574337321

CPU: https://www.jimms.fi/fi/Product/Show/166061/100-100000061wof/amd-ryzen-9-5900x-am4-3_7-ghz-12-core-boxed

MB: https://cdon.fi/kodin-elektroniikka/asus-rog-crosshair-viii-hero-wi-fi-emolevy-p55351058?utm_source=muropaketti&utm_medium=cpc&utm_campaign=muropaketti

GPU: https://www.jimms.fi/fi/Product/Show/164724/rog-strix-rtx3090-24g-gaming/asus-geforce-rtx-3090-rog-strix-naytonohjain-24gb-gddr6x or https://www.jimms.fi/fi/Product/Show/164678/tuf-rtx3080-10g-gaming/asus-geforce-rtx-3080-tuf-gaming-naytonohjain-10gb-gddr6x but will most likely wait for 3080 TI

RAM: https://www.rde.fi/products/fi/433/190256/sort/1/filter/0_0_0_0/Predator-16GB-DDR4-4600MHz-DIMM-HX446C19PB3K2-16-muistimoduuli.html?utm_source=hinta&utm_medium=referral&utm_term=hinta&utm_campaign=hinta

PSU: https://www.jimms.fi/fi/Product/Show/130026/cp-9020140-eu/corsair-1200w-hx1200-taysmodulaarinen-atx-virtalahde

M.2 SSD: https://www.verkkokauppa.com/fi/product/62107/qcjmn/Samsung-980-PRO-SSD-500-Gt-M-2-SSD-kovalevy



Cooling solution

This could be done with peltier elements https://www.radioduo.fi/tuote/moduuli:-peltierin--15-6v--30-5a--62x62x3-9mm--350w--mat:-al2o3/PM-62X62-350/

or with a water chiller https://www.aquatuning.fi/vesijaeaehdytys/jaeaehdyttimet/chiller/4742/waterchiller-hailea-ultra-2000-hc1000-1650w-cooling-capacity?c=23132

The water chiller is interesting as that would require less work from my part but I am concerned with the noise that thing makes. Peltiers are going to require some assembly that I am worried will end up looking too much DIY.

Basic diagram of the cooling system:


Then we also need:

The power of the controller will most like be not enough to run the TEC at full capacity but I am anticipating less that 30w required to run the TEC as it only needs to achieve delta C of about 5-10 degrees. This needs to be revised after the final math is done.

Some basic math needed to calculate TEC thermodynamics: http://tomswiki.wikifoundry.com/page/Peltier+(TEC)+Cooling


Unsolved issues

- [SOLVED] Materiel incompatible with the liquid. While a list of incompatible materiel can be found for the liquid, I foresee trouble in trying to figure what type of sealant they use in the aquarium and also in the pc components. Most are silicon based and that's a no go according to 3M. Better build the tank from metal.

- Passthroughs. I see 2 ways to do this. My first option is to install IO passthrough ports for every single cable to the system but if that turns out to be too difficult or too expensive the next best thing is to build a pipe from outside into the liquid, route all the cables through the pipe and use the liquid itself as an airlock. A low viscosity resin can be used in the pipe to further decrease liquid loss.

Possible products for pass throughs: https://www.thomann.de/fi/neutrik_data_liittimet.html

Nice looking gasket for some of the pass throughs: https://www.radioduo.fi/neutrik--black-sealing-gasket-d-shape-dust-and-water-resistant/p/SCDP0/

- [Unnecessary] Latches for the lid. How to attach sturdy enough latches into the glass aquarium to keep the lid shut tight?

- Cooling system automation. Ideally you would want the cooling system to ramp up as pressure builds inside the case. Measuring the cooling liquids temperature doesn't work that well with this kind of build. Temp sensors on different heights of the headspace sound like a good option to monitor the vapor levels. Powering the cooling system from the PC would also be nice but I might have to power it externally.

- [SOLVED] HDD's. I would like to submerge every component but mechanical hard drives don't like liquid of any kind. Apparently helium filled HDD's are completely sealed and "should" be fine. Fortunately I happen to have couple helium HDD's. Not much info about this with a quick googling. After searching a little it seems the hard drives are hermetically sealed with laser welding. That's enough confirmation for me, in they go!

- Fluid loss mitigation system. I would like to use bellows but I don't like the aesthetics of them. Maybe use an air balloon inside a metal shroud on top of the lid? How does the rubber of the ball react to the vapor? Most likely very badly as they have lots of plasticizers in them. DIY bellows might be the best solution.

- [SOLVED] Attaching condensator. First I was thinking of just attaching the radiator to the lid through the glass tubes with compression fittings but now I'm worried if they are enough to hold the radiator. This isn't actually an issue as a 120mm fan radiator can no way fit inside the aquarium horizontally with the mounting tray in the middle. The condensator has to be mounted vertically and then I can screw it straight into the mounting tray. The problem now is that the shroud will gather some liquid but with enough grit I'm sure I can bore some holes into the underside of the shroud....

- Coating. For reasons unknown, 3M recommends powder coating the insides of a metal case. Typical materials used in powder coating are not in the recommended list of materiel. Is powder coating really necessary? What material 3M would recommend?

-Desiccant. There are other options besides silicagel. What material 3M recommends?



  1. [Done] Learn to use CAD and come up with a first sketch of the case
  2. Contact 3M or the retailer about material compatibility
  3. [Done] Contact several aquarium makers about the custom aquarium and their sealant solutions 
  4. [Done] Contact 3M and confirm the compatibility of the sealant used in the aquarium if necessary. -> not compatible
  5. [Done] Design a second iteration of the tank but instead of using any adhesives to seal the glass panels, use bolts to attach the glass and a gasket.
  6. [Done] Design 3rd iteration of the build with suggestions given by the metal workshop.
  7. Find someone who can fabricate the lid and the frame for the case. Preferably from stainless steel. This should not be a problem as there are several metal workshops near by.
  8. Find an outlet that can provide me with the tempered glass panel according to my custom design
  9. Choose parts for the PC, get their dimensions and sum up their TDP
  10. Find someone who can deliver all the different special parts and cables needed inside the case according to material compatibility
  11. Design the cooling solution according to the TDP
  12. Design the semifinal version of the case with the correct component measurements
  13. Request quotes for all the outside work needed.
  14. Calculate costs and make a decision on execution.
  15. Order at least samples of different parts that go into the lid
  16. Design the final version of the case with the correct IO perforation measurements
  17. Order all the parts and custom work
  18. ???
  19. Profit



(very) Rough estimates:

  • Aquarium 100€ or one piece glass container could be even 800€
  • Novec liquid 2000€
  • The lid 200€
  • Bits and pieces for the cooling solution 500€
  • GPU 1300€
  • bits and pieces and RGB for the components 500€

Less so rough estimates:

  • CPU 581,89€
  • MB 410€
  • M.2 NVME SSD 145,89
  • PSU 242€
  • RAM 627,22€

Estimated total: 6607€


Edited by Drakhorn

Share this post

Link to post
Share on other sites

Leaked footage from the future, Drakhorn playing games with his new case and super OC to run........ a 16bit emulator.


Edited by Howley

Share this post

Link to post
Share on other sites

So far it has become apparent that just the price of Novec itself makes this build not the smartest one. Not mentioning the trouble one needs to go with fabricating all the parts needed for the build.

So far depending on how the the design lands, this will take about 8-11 liters of the stuff, that's about 2000-2600€ on the liquid alone. You can see why there aren't commercial solutions available to consumers, this isn't the most cost efficient way to build a pc for sure.

Still not decided whether to proceed or not. I want to see if this can be made even theoretically first.

Edited by Drakhorn

Share this post

Link to post
Share on other sites

Just a thought but what if....


What if I route a cold water line from the wall right next to my pc and plug the whole thing into mains and just flush the water after condensing?




According to legislature, cold water cannot exceed temperature of 20C, this would be perfect! Also as the loop has no micro fins I'm not worried about clogging and if I use a copper condenser and brass and soft pipes there should not be a problem with corrosion. As the loop wouldn't be closed there isn't any issues with growths in the system either.




AFAIK legislature should not prevent doing this. HOA might have an opinion about it though but will consult a professional about it first. My cold tap water measures 13C which is very nice but could impose condensation problem in the pipe perhaps? Will need to ask about it as well and perhaps insulate the pipe if necessary.



Edited by Drakhorn

Share this post

Link to post
Share on other sites
15 hours ago, Sammerkand said:

Intel Cryo Coolers.......



Just saying.

A bit off topic and frankly not nearly as "cool" as a submerged pc boiling its coolant with pipes going into the wall. Also not really an option atm for 5900X but something I would seriously consider if going for max cooling. Just to point out, Intel hasn't invented anything new here, they are basically selling you Peltier DC control software packaged in a neat hardware solution that you could build yourself anyways.


On another matter. I've hit a weird snag with ASUS customer service. Here's our conversation in terse.

Me: Can the CPU_FAN and PCH_FAN connectors be left unplugged? Is there an option in the BIOS to disable any possible fail safes for them?

ASUS: No, it is not mentioned in the manual

Me: Are you sure? I know it's not in the manual, that's why I am asking because ASUS FAQ website tells me you can disable them: https://www.asus.com/us/support/FAQ/1006064/

ASUS: Unfortunately this function is not working. I havent seen this option in BIOS in any recent board.


There are many forum posts and youtube videos showing how to disable monitoring on CPU_FAN connector in ASUS motherboards. Is this some kind of liability avoidance bullshit where they cannot claim this as a feature in case I fry my board and try to RMA it then?


Does anyone in here have an ASUS board that could verify that there is an option in BIOS under Monitor to disable the "CPU FAN speed" monitoring?

Something like this: https://www.youtube.com/watch?v=qSqnVjcEaSE

Share this post

Link to post
Share on other sites

A friend of mine confirmed that you can disable at least the cpu fan monitoring


Can't believe that ASUS either has so dishonest or just incompetent people in technical support department.

Share this post

Link to post
Share on other sites

Little update here. I haven't continued this project for 2 reasons.

1. The cost efficiency of this is not very good.

2. The cooling capacity of this solution is capped on the boiling point of the liquid used, making this technically a worse solution than just basic air cooling as the liquid would always be hotter than ambient temperature in the room. Custom water cooling solutions scale much better for cooling capacity required.

I still think about this every now and then and will consult some of my engineering friends about it so I might update and clean up the first posts in this thread but I will be building a rigid tube water cooled pc as my next project.

Share this post

Link to post
Share on other sites
Sign in to follow this  

  • Create New...