Is immersion cooling worth it? The Czech Braiins did the accounts for the antminer S19 XP.
Temperature, the miner’s enemy
Warm ambient air forces fans to spin faster and ASICs to draw more amperage to maintain hashrate.
The XP antminer is particularly sensitive to heat. For a constant hashrate, the power consumption can soar by more than 40% if the temperature of the ASICs is high. Clearly, heat drastically reduces efficiency, which is measured in watts per TH.
[L’efficience fait référence au nombre de watts nécessaires pour générer un certain nombre de hash. Un antminer produisant 1 TH pour 21 watts est plus efficient qu’un autre produisant 1 TH pour 30 watts. Notre article sur l’overclocking : Comment débrider ses Antminers ?]
The advantage of immersion cooling is that it can cool ASICs much more efficiently. However, the necessary equipment comes at a cost. So consider what is best for your own situation.
If your setup is in an extremely hot region where power is cheap, it’s a good idea to use immersion and overclock.
On the other hand, if the climate is cold and the energy relatively expensive, it is not worth it. It may even make sense to reduce the frequency of hashing ASICs to gain efficiency.
Advantages of Immersion Cooling
Before getting into the economic analysis, here are the advantages of immersion over air cooling:
– More efficient heat dissipation:
The density of the fluids used in immersion (heat transfer fluids or dielectrics) means that they absorb heat much better than air. The faster heat transfer allows heat to be removed from the ASICs much more efficiently.
– Safer Overclocking (Chop Faster):
Better heat dissipation makes overclocking ASICs safer. Up to 40% more hash or more. But, the electricity should not be too expensive, because the efficiency decreases as you chop faster.
– Increase in the life of the equipment:
Fan vibration and rapid temperature fluctuations reduce hardware life. More on this in our article: Extend the life of his Antminer.
Immersion cooling greatly reduces both of these factors. Firstly because there is no need for fans anymore. But also because the temperature of the fluid is much more stable than that of the air which varies much more (day/night, summer/winter)
That said, immersion cooling is primarily for practicing overclocking. However, chopping faster than the factory settings requires a higher voltage, which reduces the lifespan of the ASICs.
-Better operating conditions:
The heat transfer liquid prevents dust and debris from entering the equipment. This results in a considerable reduction in the need for cleaning and maintenance (and therefore in wages). Not to mention the absence of the din of the fans.
The benefits of immersion cooling are therefore numerous, regardless of the climate in which the miner operates. However, its cost must be taken into consideration.
Installation cost “air cooling” vs “immersion cooling”
Concerning air cooling, part of the costs comes from the ambient air ventilation system and the insulating materials separating the hot aisles (air outlet from the fans) from the cold aisles. Or about 10% of total infrastructure spending according to Braiins.
Most of the cost is attributable to the labor, ASICs, cabling and all the electrical equipment like the transformer.
brains assumes that a one MW (megawatt) installation costs something between $150,000 and $500,000.
Costs are almost the same in immersion (minus ~5% due to fans no longer needed). However, other equipment is added to the bill:
-Reservoirs (for soaking the ASICs in the heat transfer fluid)
-Pumps and Hoses
-Coolant
-Heat exchangers (between cold water and hot heat transfer fluid)
-Dry cooling systems to cool water
-Temperature sensors and monitoring/control systems
Braiins estimates that such equipment costs between $150,000 and $350,000 per MW (not including delivery costs). According to Braiins, it will then be necessary to count between 280,000 and 600,000 euros in all. In other words, an immersion cooling installation costs twice as much.
Comparison under different scenarios
Comparison parameters:
-BTC price: $40,000
-Period: 4 years
-Mining Difficulty: 29794407589312 (May 2022)
-Electrical power: 10MW
-5% of the electrical capacity dedicated to cooling equipment (fans, pumps, air coolers, etc.). That is 9.5 MW to mine.
-ASIC: Antminer S19j Pro 104 TH/s at $8,500 each / 3,068 watts.
-Cost of the “air cooling” infrastructure: $250,000/MW, or $2.5 million for 10 MW.
-Cost of the “immersion cooling” infrastructure: $450,000/MW, or $4.5 million for 10 MW.
-Depreciation of equipment at the rate of 20% per year (submerged miners should however have a longer operational life)
– Electricity price: $0.05/kWh
Note that Braiins did not take into account the fact that the hot summer months have a negative impact on the efficiency of Antminer in “air cooling”. The firm also assumes that the hashrate is constant, which still gives a small unrealistic advantage to the “air cooling” scenario.
Then, two scenarios were considered:
-Bullish scenario: 70% and 80% increase per year in mining difficulty and bitcoin price respectively (6.25 BTC per block + 0.25 BTC transaction fees).
-Downside scenario: 50% and 25% increase per year in mining difficulty and bitcoin price respectively (6.25 BTC per block + 0.1 BTC transaction fees)
Based on these assumptions, Braiins analyzes four different configurations:
1) Air-cooled with factory settings and firmware
2) Immersion cooling with firmware and factory settings (3068 watts)
4) Immersion cooling with Braiins OS+ firmware (4000 watts)
5) Immersion cooling with Braiins OS+ firmware (5000 watts)
Results :
-Configuration 1 – (322 PH/s)
This scenario produces good results in both scenarios. Even the bearish scenario (where the hashrate rises faster than the price of bitcoin) sports an annual rate of return (IRR on the chart) of 210%.
That said, the profit disappears if the miner has to regularly sell BTC to pay for operating expenses (electricity, wages, etc.). The miner must therefore borrow enough so that he does not have to sell a single BTC for four years.
The miner becomes profitable after 23 months in the bearish scenario.
–Configuration 2 – (338.1 PH/s)
More BTC are mined, but this is not enough to compensate for the additional investment of immersion cooling.
The miner becomes profitable after 23 months also in a bearish scenario. In other words, immersing yourself without overclocking isn’t worth it.
–Configuration 3 – (323 PH/s and overclocked to 4000 watts by antminer)
Brains writes: “An antminer S19j Pro in immersion can hash at around 136 TH/s with 4,000 watts. However, this figure is very conservative. We have seen antminers running at 140 TH/s in ideal conditions.
Since we expect to consume 4000W per machine, we won’t be buying as many antminers to fill the 9.5MW of available capacity. Instead of 3,096 miners, we will only need 2,375.”
This scenario 3 is much more profitable than the first two. The breakeven is reached after 20 months (bearish scenario):
Configuration 4 – (276.3 PH/s and overclocked to 5,500 watts by antminer)
Brains writes: “When you can safely overclock with immersion cooling, the Antminer X19s can run at 6,000, 7,000, or even 8,000 watts. However, the supplied power supplies cannot bring such power to the miner. Pushing the limits of ASICs will require purchasing new power supplies capable of delivering more than 4000W per machine. We assume a hashrate of 160 TH/s at 5,500 W (34.4 J/TH), but again, this is a conservative number. »
Due to the loss of efficiency due to the increase in power and additional costs, the miner becomes profitable after 20 months (downside scenario). There is therefore no interest in consuming so much energy.
Conclusion
Each minor corresponds to a unique situation. No two miners evolve under the same conditions. The website of brains is very handy for making projections given your settings.
Also, while overclocking S19 to over 160 TH/s sounds great, the sacrifices in terms of efficiency and investment aren’t necessarily worth it. This will mainly depend on the price of your electricity.
A few final thoughts:
-If the antminers have to be replaced approximately every five years, the investments for immersion cooling are made only once.
-Some regions are very hot in summer, which actually gives a definite advantage to immersion cooling for many months of the year.
-Once soaked in the heat transfer liquid, the antminers will not be able to be re-aired. The reason being that the liquid eats away at the thermal paste that connects the chips and the heat sinks.
If you liked this article, you will definitely enjoy this one: Can a minor solo win the jackpot?
Receive a digest of news in the world of cryptocurrencies by subscribing to our new service of daily and weekly so you don’t miss any of the essential Tremplin.io!
