[HN Gopher] Mineral Oil Cooled PC (2014)
___________________________________________________________________
Mineral Oil Cooled PC (2014)
Author : LorenDB
Score : 41 points
Date : 2025-01-24 13:57 UTC (3 days ago)
(HTM) web link (www.pugetsystems.com)
(TXT) w3m dump (www.pugetsystems.com)
| perihelions wrote:
| - _" Unfortunately our mineral oil aquarium PC kit project has
| been alleged to infringe on the patents held by a company who
| holds several patents related to mineral oil cooling of PCs."_
|
| 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"_)
| perihelions wrote:
| My comment was a bit harsh as I wrote it, but, it's genuinely
| so _depressing_ to read about the cool things people create,
| and then seeing them vanish from the internet, as morals-less
| lawyers swoop in and claim various falsehoods--on the
| calculation that the victim won 't have the resources to fight
| against professional patent trolls. Against experts who
| _specialize_ in legal extortion, such as this example.
|
| - _"...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.
| tantalor wrote:
| You could patent this, but it would be on specifics of the
| process & design, like how the chassis are constructed (down to
| the fabrication steps), special/unusual components, and exactly
| how the heat dissipation works.
|
| 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.
| atmanactive wrote:
| Been there, done that. Useless in a home/lab environment. For a
| data-center - maybe.
|
| 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.
| imglorp wrote:
| I mean, you can still get Flourinert (!!) and you can pretend
| you have a Cray.
|
| But more seriously, would distilled water work? It should have
| minimal conductivity in theory.
|
| https://www.3m.com/3M/en_US/p/d/b40045180/
| MisterTea wrote:
| I think distilled water would begin to leach metals and
| whatnot from the PCB making it conductive again. Certainly
| would need to be continually monitored and watch for
| corrosion. Flourinert is super expensive, we use it at work
| for gross leak inspection and filing a single small tank is i
| think 3-4k USD.
| qwertox wrote:
| Distilled water is pretty inexpensive, so flushing it once
| a month or so would probably work.
| dan_can_code wrote:
| I just had the funny thought of flushing my PC like a
| toilet.
| mystified5016 wrote:
| Distilled water is surprisingly corrosive. Water is actually
| a very good solvent, but most water you encounter is already
| so saturated with dissolved minerals and such that it _seems_
| mostly inert.
|
| 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.
| HPsquared wrote:
| Air conditioner condensate is not pure water though, it
| contains dissolved CO2 from the atmosphere aka carbonic
| acid.
|
| All water in contact with atmospheric air does this. Same
| with other acid gases, but CO2 is the main one. See also
| acid rain.
| madaxe_again wrote:
| Well, surely you just need a sacrificial anode - nice block
| of zinc or magnesium wired to earth that you replace once a
| year.
| simne wrote:
| > would distilled water work?
|
| 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.
| simne wrote:
| Exists sea grade of electronics, which are covered by thick
| coating layer, but it is also non-ideal solution, could
| become fractured, and you cannot make coating on dip
| switches or on connectors.
| wbl wrote:
| Fluorinert was absolutely something the USSR could produce.
| You need fluoromers and fluorine chemistry for nuclear
| technology.
| JumpCrisscross wrote:
| > _when airtight, the whole thing looses it 's thermal
| properties and starts overheating again. I guess because of
| pressure build-up_
|
| Does the oil actuallly avoid heating up that much by expanding?
| atmanactive wrote:
| It seems so. I would drill a 1mm diameter hole at the top to
| release the pressure and the temperatures would immediately
| drop. But then we are back to that horrible, horrible smell.
| I'm guessing that's why nobody sits next to powergrid AC/AC
| transformers (also submerged in mineral oil) but they are
| always secluded in a small room or left hanging high up on
| the pole. It's simply inhuman, that oil.
| ycombinatrix wrote:
| >The custom mineral oil pc project has always been intended as
| a cool conversation piece
|
| Sounds like it worked :)
| stronglikedan wrote:
| > Useless in a home/lab environment.
|
| Could you do something like a mini-split and put that half
| outside?
| atmanactive wrote:
| Yes, but then again, you could do that with any approach,
| even turbine-grade fans for loud air cooling. It kind of
| defeats the purpose.
| 542354234235 wrote:
| I think you misunderstand the purpose of immersion. You
| still have to dissipate and radiate the heat. You can't
| just do it with "any approach" because air, as a medium,
| cannot conduct as much heat. You will reach a maximum in
| air long before you reach a max in mineral oil. But it
| isn't magic. The heat has to go somewhere. That is why
| liquid coolers have radiators and why air-cooled cases have
| vents to the outside.
| atmanactive wrote:
| Yes, a stainless steel tank has thousand times more
| surface to disipate the heat away than a bare CPU. Same
| with 1-kilo heatsinks, like, for example, Noctua NH-P1
| which I am using for passive air cooling today.
| BenjiWiebe wrote:
| And a radiator the size of the tank probably has another
| 1000x better heat transfer than the tank.
| 542354234235 wrote:
| There was a project on reddit some years back where someone
| did exactly that. They had all the pumps and radiators in
| another room and ran the tubes through the wall, so it was an
| absolute monster PC running totally silent. A quick googling
| didn't turn it up so I don't know where to find it now.
| tedunangst wrote:
| LTT water cooled a whole room of workstations in a house
| some years ago. Those were some fun videos, in a horrifying
| sort of way. (And decided he wanted to do it again with his
| new house.)
| ChoGGi wrote:
| There was a post last month? The guy who built a lan party
| setup in his house. He used optical cables to reach the
| desktops from the rack.
| HPsquared wrote:
| If it was airtight and there were heat transfer problems, the
| first thing I'd think is the pressure is probably below
| atmospheric and the fluid is boiling (or cavitating the pump if
| there is one). Say, if it was filled to the top while warm,
| then sealed, the pressure would drop as the oil dropped to room
| temperature and contracted (or not, depending on differential
| rates of thermal expansion cf the container).
|
| 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.
| 542354234235 wrote:
| It sounds like they didn't circulate the fluid through
| radiators and were just relying on natural convection to
| circulate the fluid and the open tank allowed for heat from
| the fluid to be dissipated into the air. So sealing it would
| cause it to overheat immediately, same as if you saran
| wrapped an air-cooled PC.
| atmanactive wrote:
| As a matter of fact it wouldn't overheat immediately but
| some 15-20 hours later. Very strange. As soon as it wasn't
| sealed, it would stop overheating and would transfer the
| heat steadily for weeks. Keep in mind that the whole tank
| was stainless steel with some 50 liters of oil inside,
| hence the oil-to-air transfer was more than adequate.
| 542354234235 wrote:
| That makes sense, because it took 15-20 hours for all
| that oil to become fully heated and reach its carrying
| capacity. Your stainless steel tank and passive
| conduction was obviously not adequate, or you wouldn't
| have overheating problems. When it was unsealed, the air
| convection across the top of the tank carried enough heat
| that it wouldn't overheat. In another comment you say you
| use a passive cooling Noctua heatsink. If you seal up
| your PC with shrink wrap, your CPU will still overheat
| due to the lack of that same air convection. In a sealed
| system, you have to actively move liquid through
| radiators. You can move as much heat as you want into a
| heatsink, if there isn't a way to move heat quickly out
| of the heatsink, then you aren't doing anything.
| 542354234235 wrote:
| Additionally: Sealing the tank would turn the airspace on
| top into an insulator, much the same concept as a double
| pane window. It would convert the entire top surface from
| a primary way heat is moved out of the system, through
| that convective circulation of room air, to a primary way
| to trap heat in the system.
| jojobas wrote:
| Losing just the top surface would leave the steel sides
| largely in place for oil-to-air heat transfer. I'd think,
| given the odour problem, evaporation actually played a
| role in his cooling.
| atmanactive wrote:
| > 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.
| 542354234235 wrote:
| >when airtight, the whole thing looses it's thermal properties
| and starts overheating again
|
| 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.
| ZiiS wrote:
| More moden take https://immersion-cooling.thermaltake.com/
| LorenDB wrote:
| Wow, I had no idea you could get such a turnkey kit today! I
| seriously doubt I'll ever have a need for an immersion-cooled
| machine, but I will definitely dream of the cool setups that
| could be built in this case.
| bruce343434 wrote:
| I really doubt that the fan motors in typical PC components,
| designed to move air, were made to endure the relatively
| extreme viscosity of oil.
| ZiiS wrote:
| At this point you are running thousands of dollars of
| sensitive electronics outside their design and testing whilst
| voiding your warrenty. Wearing out a few fans is probably
| tolerable.
| yellowapple wrote:
| The authors of the article report that the fans ran just fine
| in oil, contrary to that same assumption.
| indrora wrote:
| 3M has a whole paper [0] on their use of NOVEC for whole
| datacenter cooling and had a really cool demo [1] of it a
| decade ago
|
| [0] https://multimedia.3m.com/mws/media/1838765O/3m-fluids-
| for-d... [1] https://www.cablinginstall.com/data-
| center/article/16477494/...
| hinkley wrote:
| I used to have that tank.
|
| Biowheel filtration is the only way to go.
| chris_va wrote:
| If you really want to have fun, use ln2 (and make sure you have a
| well ventilated room, unless you want to asphyxiate).
| hypercube33 wrote:
| My buddy and I did this with a $1/ gallon fish tank, 8 gallons of
| horse laxative (heavy mineral oil - real cheap way to buy it in
| bulk) and a spare gaming pcs guts. it worked but wasn't impressed
| after it heat saturates - you definitely need to cool the oil
| after say 8 to 12 hours otherwise it probably is worse than air
| cooling.
|
| the other downside is you absolutely need to dishwasher anything
| that was submerged and it makes one hell of a mess.
| hadlock wrote:
| We did this in the late 90s with an old 486 to see if it
| actually worked. That's the earliest I recall seeing this. We
| played doom on it. As you note, it created a giant mess and
| pretty much immediately went in the trash as soon as we'd
| proven it worked.
| bitwize wrote:
| Fluorinert was used as a coolant in some 80s/90s computers,
| including one of the first SFF PCs, the Ergo Brick.
___________________________________________________________________
(page generated 2025-01-27 23:02 UTC)