Mineral Oil Cooled PC (2014)
pugetsystems.com
pugetsystems.com
In my experience, the oil is far from odourless. It seems that way at first, but after a while it saturates the room/house with very annoying smell. Several months later, and I had to get rid of that smell. Even went as far to build stainless steel tanks with gaskets used for car motors. But, when airtight, the whole thing looses it's thermal properties and starts overheating again. I guess because of pressure build-up. So, it is either endure the mineral oil vapors, or nothing.
On top of all that, pray you never need to replace/upgrade a component... getting the whole thing out of the oil and cleaning it up so you could handle it again - a nightmare.
But more seriously, would distilled water work? It should have minimal conductivity in theory.
Distilled water would work for a while, but will literally dissolve the metals on your board. Conductivity will of course creep up over time, but your main concern should be that the widget you're trying to cool is getting eaten by the coolant.
As an example, you may have noticed that air conditioner condensate can damage concrete. The condensate is pretty pure water, and physics really wants it to be less pure. The water is so reactive that it can rip the minerals out of cement, leaving just the aggregate behind.
If all that weren't bad enough, once you put a PCB in pure water and it starts leeching ions, you suddenly have an electrolysis bath with dissimilar metals. If you can't maintain the purity of the solution, you'll have a very bad time.
All water in contact with atmospheric air does this. Same with other acid gases, but CO2 is the main one. See also acid rain.
https://www.sciencenewslab.com/this-beautiful-golden-chamber...
>In order for the light from these shockwaves to reach the sensors, the water has to be cleaner than you can possibly imagine. Super-K is constantly filtering and re-purifying it, and even blasts it with UV light to kill off any bacteria.
>Which actually makes it pretty creepy.
>“Water that’s ultra-pure is waiting to dissolve stuff into it,” said Uchida. “Pure water is very, very nasty stuff. It has the features of an acid and an alkaline.”
>“If you went for a soak in this ultra-pure Super-K water you would get quite a bit of exfoliation,” said Wascko. “Whether you want it or not.”
>When Super-K needs maintenance, researchers need to go out on rubber dinghies (see above) to fix and replace the sensors.
>Matthew Malek, of the University of Sheffield, and two others were doing maintenance from a dinghy back when he was a PhD student.
>At the end of the day’s work, the gondola that normally takes the physicists in and out of the tank was broken, so he and two others had to sit tight for a while. They kicked back in their boats, shooting the breeze.
>“What I didn’t realise, as we were laying back in these boats and talking is that a little bit of my hair, probably no more than three centimeters, was dipped in the water,” Malek told Business Insider.
>As they were draining the water out of Super-K at the time, Malek didn’t worry about contaminating it. But when he awoke at 3 am the next morning, he had an awful realisation.
>“I got up at 3 o’clock in the morning with the itchiest scalp I have ever had in my entire life,” he said. “Itchier than having chickenpox as a child. It was so itchy I just couldn’t sleep.”
>He realised that the water had leeched his hair’s nutrients out through the tips, and that this nutrient deficiency had worked its way up to his scalp. He quickly jumped in the shower and spent half an hour vigorously conditioning his hair.
>Another tale comes from Wascko, who heard that in 2000 when the tank had been fully drained, researchers found the outline of a wrench at the bottom of it.
>“Apparently somebody had left a wrench there when they filled it in 1995,” he said. “When they drained it in 2000 the wrench had dissolved.”
I wonder if the reporter didn't catch on and he actually had his head hanging in the water (so the water was touching his scalp).
Also the "even blasts it with UV light" isn't actually that weird. We have a well and we treat the well water with a UV light disinfector. What's stranger is why you'd need to worry about bacteria growing in ultra-pure water. What would the bacteria eat?
Unfortunately, not, because water is very corrosive, and easy gather many non-inert substances just from air.
This is very serious problem. I once talked with old support engineer from USSR, who have some experience with old water-cooled computers.
And he said, it was just nightmare, all those pipes periodically leak (some because rot, others become fractured) and in many cases, if water reach pcb, its broke.
Really problem was solved on Cray-2 (and later models) by use Fluorinert, which was so inert that was safe to literally drop boards into liquid.
Some persons even thinking, Soviets Elbrus was not successful just because USSR don't have companies like DuPont or 3M to produce something similar to Fluorinert.
May be. But in fact - not produced. And Soviet Elbrus computers used water cooling with extreme problems.
- I'm 100% sure, if it was possible for Elbrus designers to use Fluorinert, it would be used, but it was impossible, so not used.
Does the oil actuallly avoid heating up that much by expanding?
Older ones might use toxic materials like PCB though.
Sounds like it worked :)
Could you do something like a mini-split and put that half outside?
Some air space at the top of the sealed container (preferably dry air or nitrogen) would fix that. Make sure there's a pressure gauge and some means of preventing overpressure, though.
Sure, if this was my data-center investment job. For home use, I simply switched to passive air cooling.
Were you pumping the fluid through radiators? It doesn’t really sound like you were, which would be the obvious reason it would overheat. You have to actually transfer and dissipate the heat. What makes the immersion cooling so good is that it conducts heat so much better than air, you can add giant radiators and dissipate an ungodly amount of heat, so can run your equipment ungodly fast. You can’t just drop it into an aquarium of mineral oil and expect it to do much.
I never understood the over-clocking obsession. Wait 6 months and the new hardware coming out will run faster than your mineral oil computer and cost the same as it always has.
b) buying a lower cost model and getting similar performance to what you get from a higher cost model out of the box is a good value
c) sometimes you really need more performance, so nothing else will do
d) modern overclocking is more about tuning voltages lower, but not too low, so you generate less heat and can clock a bit higher, which can end up saving energy.
the other downside is you absolutely need to dishwasher anything that was submerged and it makes one hell of a mess.
The smell was not an issue for me, I put the entire contraption onto a balcony and used a long VGA cable.
The oil likelu provides better heat disipation for all parts of the fan as well, which could also prolong life.
Otherwise, with better disipation, you can probably remove fans from many components, possibly swapping heatsinks in some cases.
[0] https://multimedia.3m.com/mws/media/1838765O/3m-fluids-for-d... [1] https://www.cablinginstall.com/data-center/article/16477494/...
It should be available for wide-spread commercialization, patent-wise.
Biowheel filtration is the only way to go.
Still it's cool to see people making it work
What risible clowns. How could you dream of patenting this—oil-immersion cooling, in computing, goes back *at least* to 1961 [0] (and there's probably even earlier prior art, I can't find citations for the vacuum-tube era but it'd be very likely they had tried something like that). It's the most obvious solution in the world. "Oil-dunk hot thing for more cold" is a bronze-age blacksmith's wisdom.
[0] https://en.wikipedia.org/wiki/IBM_7030_Stretch ("The memory was immersion oil-heated/cooled to stabilize its operating characteristics")
- "...regardless of how we felt about the validity of this patent pressure, we could not devote the time, finances and energy to take up that battle"
These types of predators are the natural enemy of HN.
It should be basically impossible for some hackers in a garage to accidentally infringe on a patent without reading it. If it was infringing, it would be 1) extremely unlucky, or more likely 2) the patent was fatally vague to begin with, and should be invalidated.