Anyone doing any Asic immursion cooling?
I’ve drafted up plans for an immersion cooling system, I’d love to discuss it with someone in person.
After a lot of study I find that immersion cooling is not all that difficult, and actually has been an integral part of ALL of our daily lives. Transformers are submersed in dielectric fluid to keep cool. That’s one of the most important parts of our power grid. Submersing electronics in non-conductive fluid is the easy part. The more difficult part is a matter of how to most efficiently pull heat (thermal energy) out of the fluid. I’ve designed a system of conductive non corrosive pipes and pumps to move the fluid to a place that the heat can be dumped. The specific way that’s done is would depend on your location, how much space you have available, resources and geography also play into being able to assemble a low cost solution.
Of course you could always go to a place like GRC cooling and pay small sum of $63,000 for a single rack capable of hold just a dozen asics, which would be more than the hardware itself. Frankly pricing is ridiculous out there, the only way to make it workable is to do it yourself.
I find immersion cooling very interesting but I am wondering if it is really that much effective.
What I’ve seen in some videos is that you of course remove the fans from the miner. But do the fans really take so much power in relation to the computing power?
Also for the immersion system you need a good pump to keep the liquid in motion and some strong large fans on the outside unit. I gues that these do take also a lot of power, or don’t they?
I’m right there with you Aaron. So I’m interested what you come with. I agree that I don’t think it’d be hard to diy a working system for far less $ than the few commercial systems designed for at home mining. My biggest issue is locating the heat exchanger/fan system. The layout of my home/garage plus being in an HOA neighborhood doesn’t lend itself to an external heat exchanger. For the tank, I figured I’d either weld it up from aluminum sheet or maybe buy an old ice chest off of Craigslist, which likely would be cheaper. I’d be looking to make setup that can handle 4 newer ASICs.
There is a lot of talk about efficiency and less wear and tear on the hardware. I look forward to more DYI info on this subject.
A 120mm fan that spins to 6000 rpm draws about 2.7A @12V. P = V x I, so 32.4W. If you have 4 fans per ASIC, then it’s roughly 130Wh per ASIC, or 0.13kWh/ASIC. Assuming I did my math right, which is a big assumption, then at $0.11/kWh for electricity, it’d be just under $0.35/ASIC/day (0.13kWh x $0.11/kWr x 24 hrs) - at full amp draw. So not much, but not zero either. One would have to calculate the cost to buy or build the requisite number of fan emulators (if the ASIC won’t run with the fans disconnected) vs just leaving the fans installed and spinning through the dielectric fluid. It wouldn’t appear that it’d take long for a fan emulator circuit to pay for itself - if needed.
Do you have a well? Or free water? Use a tank of water as a heat exchanger. A thermometer attached to a valve would drain the water when its get too hot. When water falls below a certain level another valve opens to fill the tank.(The same mechanism thats used to fill your toilet bowl). If you have to pay for water its probably not a cost effective method. I also live in a colder climate which only makes such a method even more effective.
Edit: dielectric coolant collect dirt easily, any pumps running NEED a good filter to keep the coolant clean, this is why its advised to clean your hardware before submersion
Biggest hurdle for me is the cost of coolant. Its around 2500-$3000 for a 55gallon drum.
Also, do you know where to get fan emulators? I haven’t had any luck in finding those yet.
I don’t, unfortunately. I do have a 20k gal pool but live in FL so it’s not much of a heat sink except in January. It wouldn’t take long for that pool to get uncomfortably warm. It’s not unusual for water temp to hit 90F during the summer.
Good question regarding fan emulators. I’ve seen a few emulators offered online for some of the older ASICs (mainly Antminers). But I have Goldshell ASICs and I haven’t tried to run them (from cold for merely a few seconds) if the fans are disconnected. I assume that Goldshell would’ve built in some redundant fail safes (i.e. thermal sensor, fan failure detection)
ok, so far I can follow you.
The power for 4 fans of a miner is about 3,12 KW per 24h.
I googled for a pump that is indicated to be good for oil liquids. A randomly choosen one that takes 1,2KW - so 28,8KW per 24h.
If I dont have a well or free water as mentioned by @Aaron_Raycove then I need at least one large cooling fan at the outside running. Not considdering the reall need of the size and specs I google for a spare Motor for a aircondition outside unit. Just to have a number. It says about 85W - so 2,04KW per 24h.
So lets say the immersion system itself take about 30,5KW a day.
Devided by the power from a miner 30,5 / 3,12 means I need to run at least more then 10 miners in an immersion system to get any benefit in terms of power consumtion.
Is that the correct way of calculation?
That would be correct, a proper system won’t use oil though. The proper dielectric fluid flows like water, so a standard water pump is more than enough to move it. I believe a small pump 1gpm, would use up 10kw a day instead of the 28 required by a more powerful pump.
Aaron - Safe to assume that you’ve already discovered these guys?
Their white paper would be super helpful for a diy’er.
Here is my crude drawing of an expandable Immersion cooling system. It would use a central reservoir to house the coolant and circulate across the tanks the hardware is submerged in. The reservoir would dump heat into the attached heatsinks.
I have heard of them, their fluid is very expensive. Castrol sells the same fluid for much cheaper. Its been a while since I saw them, $55 for their papers. Could be worth it.
This is a great subject!
I agree, keep the content coming!
In your drawing, what are the heat sinks? Are you going to use an air-liquid heat exchanger? Liquid-liquid heat exchanger? The diy’er el cheapo plan that I have in my head (so take that for what it’s worth) was to use a rectangular tank (with hinged cover . . or an old freezer chest) to house ASICs and most of the dielectric fluid, pump (probably 1/2 hp hot tub pump), and air-liquid heat exchanger w/fan(s). I didn’t see the need for a reservoir though considered it just to reduce the chance that the pump would draw air instead of fluid.
I know that Engineered Fluid’s Bitcool is very expensive compared to using rather than using admittedly cheap mineral oil - and flies in the face of my el cheapo plan as Bitcool would become the single costliest item in the entire build. However, mineral oil has weaker heat transfer properties compared to Bitcool, higher viscosity (so more strain on equivalent sized pump), and can smell a bit if it gets too warm. Plus I’ve heard that mineral oil can cause eventual problems over time with some ASICs’ capacitors, though I don’t know if true or not
Does Castrol have a similar product to Bitcool or are they just rebranding commercially pure mineral oil?
When I was fiddle fu@$ing with my car back in the day, I was told that water is the best substance to convey heat. Would this not be true with “Emersion Mining”. Why do you need oil? Also glycol is pretty good as well, in other words why mineral oil.?
ever dropped a toaster in the bath tub?
The reservoir isn’t necessary, is was more a part of a modular design to make expansion easier.
The idea was a liquid-liquid exchanger utilizing thermally conductive plastic pipe. Really cheap compared to copper, and won’t corrode as easily. I have access to a creek to dump warm water into and well to draw new COLD 50F water from. And in the winter it’ll be even colder.
I was going to built a prototype tank out of plastic sheeting, caulk, and plywood. Total estimate for the project at about $5000, capable of handling a lot of heat load.
I’ve been working on a smaller set up as an experiment since I don’t pay for electricity in my office. There isn’t a lot of information out there for smaller scale (or really any for that matter) so I just decided to jump in head first and post the results. I was posting on Reddit but will put the links to them here:
2 Likes