Need help for Immersion Cooling project!

[Endikadija]

Hi to all,

I am starting an experimental project done by myself to cool 45 Antminer S9's through Immersion Cooling. I have created the tank, bought a DryCooler 100KW together with the Heat Exchanger and have connected to the tank.

Now the problem is at the liquid. Based in my calcuation regarding to flow and the liquid needed for miner I need around 600 liters of liquid.

I have found a liquid used in transformers with the price of 1.5 USD for liter(far cheaper than the companies that sell online liquid cooling) named STO-50 SILICONE TRANSFORMER OIL  which is made 100% Polydimethylsiloxane silicone oil with a viscosity of 50cSt (centistoke) @ 25°C. More Specs can be found at http://www.clearcoproducts.com/silicone_transformer_oil.html

Does anyone has information if it can be used as a coolant for ASICS?

[1715611801]

1715611801

[1715611801]

1715611801

[1715611801]

1715611801

[NotFuzzyWarm]

That should be a perfect choice and in fact is one that I've recommended in the several existing threads here that deal with immersion cooling. It has 2 primary advantages over mineral oil:
 1) It does not attack plastics or epoxies - only warning is to NOT use any kind of silicone glue or sealant as it WILL attack them and turn them into a crumbly mass of rubber.
 2) It is a fairly thin fluid meaning it pumps and flows very well.

Don't know what temperatures you used to calculate the dry cooler capacity being able to handle 100kw but it may be able to handle more power. Of course it depends on what your highest outside air temp can be but on the miner side, something most folks forget to account for is that miners like a VERY warm chip temperature. As in around 70-degC or more. That translates into you being able to run with rather high oil temp, you probably find that a fluid temp of 50C will work perfect. If you've done the thermal calculations using a lower target oil temp -- good news! You can dump more power in. 

[Endikadija]

That should be a perfect choice and in fact is one that I've recommended in the several existing threads here that deal with immersion cooling. It has 2 primary advantages over mineral oil:
 1) It does not attack plastics or epoxies - only warning is to NOT use any kind of silicone glue or sealant as it WILL attack them and turn them into a crumbly mass of rubber.
 2) It is a fairly thin fluid meaning it pumps and flows very well.

Don't know what temperatures you used to calculate the dry cooler capacity being able to handle 100kw but it may be able to handle more power. Of course it depends on what your highest outside air temp can be but on the miner side, something most folks forget to account for is that miners like a VERY warm chip temperature. As in around 70-degC or more. That translates into you being able to run with rather high oil temp, you probably find that a fluid temp of 50C will work perfect.

Thank you for the information, frankly I was afraid to use it since the transformers has more 'not so delicate' components and I thought it will calcify the hash boards.

Here in Albania the temperature can go up to 38 Celcius Degree in August. I think that the DryCooler can handle more than 45 S9, I think around 75-80 pieces.

Edit: My first plan it was that the cooled liquid to go in the tank at around 20-25 Celcius.

[NotFuzzyWarm]

Boards and circuitry will be fine. That fluid has been used used in many makes of solid-state HV power supplies for decades and I've never seen any issues from using it.

20-25C will actually probably be too cold and the miners will not be happy until the chips get warmer. Again, you will want to look at reported chip temps - not the 'air' temp in and out. With immersion cooling the chips run much closer to fluid temp because heat transfer between the chips and fluid is much higher as opposed to air cooling where chip temps are much higher than the cooling air.

[Endikadija]

Boards and circuitry will be fine. That fluid has been used used in many makes of solid-state HV power supplies for decades and I've never seen any issues from using it.

20-25C will actually probably be too cold and the miners will not be happy until the chips get warmer. Again, you will want to look at reported chip temps - not the 'air' temp in and out. With immersion cooling the chips run much closer to fluid temp because heat transfer between the chips and fluid is much higher as opposed to air cooling where chip temps are much higher than the cooling air.

So you suggest that the input fluid should be around 50 Celcius Degree?  Do you have any clue that how fast can the miners heat the liquid up to let's say 80 Degree?

[NotFuzzyWarm]

Mainly depends on how much fluid is in the system.
Refer to https://sciencing.com/calculate-change-temperature-2696.html for calculating temperature rise.
Per Clearco's data sheet http://www.clearcoproducts.com/pdf/heat-transfer-fluids/NP-PSF-50cSt%20Silicone%20Heat%20Transfer%20Fluid.pdf the fluid has a specific heat of 1.5 J/g °C and has a specific gravity (density) of 0.960 g/cc

As for power put into the system, 1 J=1 Watt-second. Since a s9 typically pulls around 1,350W that gives a power input of 1,350 Joules per-second per miner. Oh, don't forget to add in the heat load that your pumps add to the system: Quick and dirty answer to that is it will be equal to the power used by the pump(s).

You mentioned using 600 liters of fluid and of course 1 liter=1000cc. So to start you out: 600,000cc's * 0.96 = 576,000 grams of fluid
Have fun with the math

[freak1]

Mind sharing what kind of pumps and tanks you are using?