What’s the max PSU overload for immersion cooling?

I’ve got my S19j Pro running at 104TH with some upgraded air cooling, and I've cranked it up to 121TH while pulling about 3.5kW. That’s around 15% over the specs (3068 W), but my PSU seems to handle it well with its 12-15V and 233A range. So, here’s my question. What if I pushed it to 4kW or even 5kW? I know more current brings more heat, but let’s say I have solid immersion cooling to keep things chill. At what point will the PSU give up on me, and what factors lead to that failure?

Most PSU's are rated to 100% load at up to 40C and de-rate as the temperature increases. So if you reduce the operating temperature you can drive them a little harder than their rated capacity. You may well get away with 110-120% of rated load with adequate cooling. However you're talking about >30% overload, it is unlikely that the components used in the power supply will be sufficiently over-rated to cope with such a heavy overload condition, if there's any integrated protection you'll likely run into that first and the PSU will probably just fail to start or shut-down as the protection takes effect. If its a poorly designed power supply with little or no protection features you'll likely just destroy it by overloading so heavily even with extreme cooling, worst case you'll start a fire. I hope your insurance is up to date, though if there was a fire caused by the overload you would likely be considered negligent and you might find your insurance wouldn't cover you. In all respects overloading a high capacity power supply is a VERY BAD idea.

Thank you for your response, but it is too generalized to answer the question. Lets dig a little deeper, what electronic components will fail first and why? Are there any components inside the PSU with some kind of insulation, so that external cooling will not help? Are there any components with electron diffusion and/or rapid material aging, which will finally lead to failure? Are the some side effects regarding the software or control units? Are there any other harmful side effects i do not know yet?

Just get the right psu they are available https://www.ebay.com/itm/295979681986 I have a few of these if you want one. you will not be able to pull too much power on these or get one on ebay direct https://www.ebay.com/itm/365136712946? seller has 2600 deals all positive

Given your response, I guess you don't know an awful lot about electronics or power supply design. I've been designing off-line switching power supplies for around 30 years. You're making the assumptions that the original designers have allowed an overhead in their design of more than 30%, in general this could not be further from the truth. Most of the time designs are made to a price. So components tend to be heaviliy optimised for a particular operating condition, exceeding this will result in failures and these failures are not all down to operating temperature. It is beyond the scope of a reply to teach you how switching power supplies work and what the weak points are in them. However a simplified list would be A basic PSU design can be broken down into 4 sections 1) Primary rectification : Consisting of diodes or full FET (transistor, somtimes called an ideal diode) bridge, both approaches will be designed for a particular maximum operating current. Exceeding the specifications of the parts used will result in failure for diode and FET rectification they tend to fail short resulting in the primary fuse going open circuit (assuming the design has one), however in failing they often damage PCB and surrounding components. 2) Power factor correction: Following the primary rectification given the operating power, there will be some form of power factor correction circuit, this will have one or more switching elements consisting of FET's and rectification diodes or other rectification FET, the FET's will again have both voltage and current limits which when exceeded will cause permanent damage to them. 3) Depending on the swiching supply topology and for the kind of power required it is likely a full bridge forward converter or resonant LLC circuit, both of which are heavily complex designs having multiple FET switching elements, snubber circuits and output conditioning and smoothing all designed to meet a particular current and power requirement, you exceed the power design and permenent damage will be done, again not necesarily because of heat, but because you have exceeded the power delivery capablilities of the components in the design. 4) Control loop elements, operational amplifiers, opto-couplers, fixed and variable voltage references, protection and monitoring devices and the switching controller. These elements are less likely to fail but may fail as a result of parts in the power path failing short and delivering high voltages where they do not belong. Hopefully if the design has been done properly there will be over current and over voltage detection at each stage of the power supply in order to prevent you from destroying the supply by overload, short circuit or by overvoltage. Damage to components does not come purely from generated heat, although it is a significant factor in the failure of parts due to thermal run-away. Thermal run-away occurs where a part is driven beyond its operating limits and it starts to fail in such a way that it becomes resistive or its leakage current exeeds the devices capability, at which point they let out the magic smoke. So all in all I reitterate what I said. Overloading a high capacity power supply is A REALLY BAD IDEA!

That is the answer i hoped for. Not regarding the result but the expertise and deeper explanation. Thank you very much. And this seems to be the right solution. I already stumbled over the website of alphaminer but i did not noticed any prices there and did not dig deeper. I guess i am going to look for some distributors in EU, even if I have no specific need for now.

You're welcome. I would suggest you avoid using unbranded power supplies, especailly those that do not carry safety marks on them CE, FCC, UL, NEMCO, TUV etc etc. With unbranded PSU's you have zero guarantee that they pass even basic safety standards, given the operating powers can you really risk that? If you decide to use one of those Alpha Miner power supplies, please ask the vendor for a copy of the safety approvals, even if they're just an importer they are responsible for ensuring product safety and regulatory compliance. Safety first!

if you are doing immersion. I would do this buy a blow torch go out doors on a concrete pad or a sandy area. get a cup of your coolant (does not apply if you used distilled waterblock system ) put the cup of coolant in a stainless steel  bowl fire up the blow torch and see if the coolant burns. some coolants are flammable and you do not want to use them to cool off 5000 watts of psu and hashboards. also test the coolant to see if it conducts electrical current on low levels it should not.

Good point! The correct kind of liquid coolants for immersion cooling are typically PFPE (Perfluoropolyethers) such as Solvey Galden, which has a high dielectric strength (i.e. non conductive) is non flammable and non toxic. Interesing side note, Galden PFPE is also used in vapour phase soldering for electronics. IIRC its expensive depending on the temperature requirements.

I think destilled water is only relevant for watercooling, because the effort to keep it dielectic is too high, right? Regarding the PSU the 92% Efficiency of Alphaminer PSU tells me that heat development of PSU (8%) is much less then heat development of Hashboards (92%). Is watercooling meaningful for PSU, due to "centralized" heat sources (like asics at the hashboards) or is it unsuitable? Regarding coolant of immersion cooling i guess safety effort is a tradoff for other requierements. Obviously non flammable coolant is desirable, but i am not sure if one should take it absolute. For example pressurized Propan and Butan are coolants for heat pumps, low climate impact, low price, dielectric and good thermal characteristics, but flamable. I guess no one ever tried that as a coolant for mining hardware (i will neither), but i would not rule out that, because with highter safety effort the risk seems to be manageable.

With 92% efficiency that equates to 80Watts dissipation per KW of load, so at 4KW load the PSU will be dissipating 320Watts of heat. This is not insignificant and will definately need good thermal managment. I think the Alpha Miner PSU's are designed for immersion cooling, though they do seem to list forced air cooled models as well. Using anything flammable as a coolant is asking for trouble. Butane/Propane when used as a coolant is used as a non-contact refrigerant much like R134a air-con refrigerant and is typically passed through a compressor into a heat exchanger. It is not used as a direct contact coolant like with immersion cooling.

Flammability aside, directly using Butane/Propane is out of the question due to the pressure needed (145-200psi) to liquefy it and the fact that the immersion tank needs access for the miners so not only big but also needs a gas-tight hatch. Better to use a Flourinert that is liquid with a boiling point of say 140F and condense the vapor back into the tanks using water from a dry cooler. That is how the BitFury farms work. ref https://liquidstack.com/2-phase-immersion-cooling For AI data center needs they have https://liquidstack.com/cdu-direct-to-chip/coolant-distribution-unit-1-megawatt

Good point. I am not sure if it would be possible with hydro miners but i guess the pressure might be too hight for them. Btw: do you know some good hydro cooled overclocking PSU for hydro miner?! @notfuzzywarm: i don't think 2-phase immersion cooling is the best solution. Regarding heat flow only it might be the best solution but regarding the economics they are not. as far as i know the 2-phase coolants are much more expensive than other coolants and not nessecary with current heat flux density. In Addition to that the 2-phase coolants i know are mostly harmful to the climate and therefore not easy to get in europe.