I know the comments are old, but I think what he meant is, that when there is extreme boiling, the bubbles of air will use too much espace "arround" the chip surface, and that air would prevent the liquid to make contact with the chip surface. That would in theory reduce the efficiency.... While in the other hand, a bigger heatsink (well seated on the chip) would give a bigger surface for the liquid to make contact and to dissipate the heat.
I will agree somewhat on the "no practical cooling advantages". For a desktop system, it is not cost effective, and completely not practical compared to air cooling. It's just interesting. If you want practical, read down in the thread about computer data centers. Imagine 100 blade PCs lined up in a Novec bath, spaced a few mm's apart. No heatsinks, no energy wasted on fans. Vapor condensed by conventional water radiator, that carries heat outside. Now, no HVAC systems are needed. BIG SAVINGS
@kristofferD 1) It's not about transferring the heat to the liquid (as is the case with oil cooling, or water cooling) With this, it is about the phase change from liquid to gas that transfers the heat. The liquid has a very low boil point, so it cools only the areas (bubbles) that have greater heat than the liquid's boiling point. 2) No the liquid evaporates in seconds, and the board is 100% dry to the touch, nothing remains behind, no residue, no oils, no dust, no rust, nothing
@TheScyy, we did not test video cards... but there is no reason it would not work. Tests have been done at 3M, where only 200cc of this fluid (the version with a 49C boiling point) was able to continuously cool over 4000watts. ( +/- equivalent of (30) i7940 CPUs @ 100% load). Remember this isn't magic, you still have to do something with all the heat. (increase the capacity of the TE coolers on the top, or switch to a liquid cooled radiator to move the heat out and condense the vapor.)
@backstyck : I suppose it could get so hot that it produces a layer of gas around the chip, but the expanding gas leaves (rises) and is replaced, so cooling still occurs even if the fluid does not directly contact the chip. It's the phase change, the actual act of boiling, *not* direct heat transfer to the liquid fluid, that makes this particular process so effective. Cooling the liquid is a waste of energy, you need to condense the vapor to complete the phase change process.
@Uber1nsane, I'll see if I can find out and post it. We're an industrial computer system manufacturer, and got it directly from 3M. Depending on the "flavor" ~$300 per gallon give or take. That's why we used the foam (it was to displace enough liquid so that 1 gallon would be enough) The fluid can dissipate over 4000 watts of heat with only 200cc of fluid. People serious about using this stuff, have to design computer hardware to be very dense, then it's crazy efficient compared to fans.
@thebeetalls : Not sure. We were watching the temp. of the liquid, & the temp that the CPU was reporting. By the nature of the fluid, it boils at a specific temperature. It's the boiling that removes heat, provided the fluid is replaced (because it's submerged) this limits the "allowed" heat of any part on the board. Treating the CPU surface with a chemical boiling plate, will greatly improve the cooling delta. It's not about increasing the surface area, but rather ensuring boiling across it.
@CaptainDishman We didn't overclock. However in theory, the temps should remain constant. The liquid always boils at the same temp, and should maintain low temps indefinitely provided the vapor is being condensed quickly enough that you don't run out of fluid.
@PortaleMedia , It was running "BurnInTest v6.0" You're right though, "Speed Fan" was installed. Can't easily run the tests using other software, we already disassembled that rig. Hopefully I'll be building a version soon, that is *much* more interesting than the proof of concept. We monitored the various temperatures, as well as the fluid temperature (the multimeter in the video, showing Fahrenheit temp. from a probe) the whole time. It's was all good.
@crabid yes, see lower thread comments:TEC is compact but inefficient. Other methods are possible, even 100% passive w/enough heat sink surface area. You just have to get the heat out somehow (condensing the vapor) 2nd: You're thinking water. here, boiling = cooling. CPU is always 5-18C over boiling (varies depending on CPU surface treatment used to encourage efficient boiling) It can't get hotter. 3M did a successful test cooling 4000w, w/only 200ml of fluid. (same as ~30 i7940's @ 100% load)
@LechuCzechu Sort of. Allowing the container to pressurize and then trying to regulate that pressure would greatly increase the complexity. Venting the pressure = fluid loss. Fluid = ~$300gal. The fun thing with this is that is temp is regulated by boiling point, and the whole board, all parts have equal cooling.
@vipondiu, There are different types of Novec fluid, some with higher boiling points. Novec 649 boils at 49C, and will be what we use if we revisit this project. With a good boiling surface on the CPU, the chip core temp will be somewhere around 55 to 62C. Hard drives are sealed from dust, but not atmospheric pressure. They all have vent holes to equalize the pressure. Those can be sealed, and the hard drives submerged, but no guarantees on disk life, performance or warranty after that. :-)
@crabid No problem :-) Even with a gas layer, consider how quickly red hot metal cools in water, just a few seconds for the boiling to stop (~600C to >100C in almost an instant) then consider how much longer the metal remains hot to touch (when not boiling, the rate of cooling is much slower) Now consider that this liquid (depending on the flavor) boils at ~1/2 to well under 1/2 the temp of water. That means this very aggressive cooling is happening at, and maintaining, much lower temps.
@otoluk In our test, the CPU was not treated to encourage even, complete boiling, so the numbers are not as good. There is a special paint that can be applied to the surface, or, a boiling plate can be bonded to the chip. (Not to add surface area, but to provide a good boiling surface) If memory serves, when that is done the temp of the CPU compared to the temp of the liquid boiling point changes to somewhere between 2 and 5 degrees over the fluid boiling temp. (depending on treatment method)
Followup: Talked to 3M, and found out that the oils in the flexible PVC plastics are needed only during manufacturing, and do not really affect the performance of the plastic. So while it leeches the oils out of the plastics over the long-term, it does not cause them to fail or become brittle. At the worst, you'll end up with some oil droplets condensing on the walls. (placing activated carbon (a fishtank filter) in the liquid, will absorb the waste oil if this is a problem)
yes. If you click the "sort by thread (beta)" link, it will organize these comments into a conversation that can be read where a lot of questions are answered. Cost wise, if you are not phase changing, some other substance would be better that doesn't evaporate or cost so much. We did not test with graphics cards or overclocking. Just a proof of concept. :-)
Boiling is a method of cooling or maintaining temperatures. If a liquid will boil at a given temperature, exceeding that temperature converts the liquid to a gas. When you pour water on a hot pan, and the water turns to steam, it cools the pan... very rapidly. Similarly, this liquid boils at a very low temperature, so it will turn to vapor much more quickly, thus removing the heat from the hot surfaces. The difference is, this liquid does not conduct electricity, so it is harmless to a PC.
Our engineering department is a bit high on ambient noise. (various fans, beeping, etc.) But generally speaking, it's dead silent, particularly with the cover on it. All components could be submerged, but one has to deactivate speed sensors. The heat still has to be processed out, so there may be noise depending on your chosen cooling method. Unless you use something like a remote water pump to cool the condenser heat sink, or a larger passive heatsink (with higher boiling point liquid).
@backstyck 3M cooled 4000watts of heat in an 8x8" area, w/200ml of fluid. (~equivalent of 30, i7940's) Remember, it's not heat to liquid (like in mineral oil cooling). You're vaporizing the liquid to move the heat, it's the phase change that does it. I guess jets might help if it's just crazy hot, but a larger boiling plate (not a heat sink) would probably help more. The idea is to instantly boil the heat away before it get's hot enough to create the massive cavitation bubbles.
@mranenome, some of the sealed "Heat Pipe" technology uses a similar principal. The "magic" here, comes from the phase change. Most conventional computer liquid cooling systems are designed for the liquid to remain in liquid form, and would not deal well with pressure from the expanding gas.
Exactly. It limits the temperature of any part of the board... it's like a giant super-effective heat sink covering every component from every side. The temperature delta depends on how effectively the surface of the component allows boiling. If we do it again, we'll probably go with a fluid with a slightly higher boiling point, so we can more readily condense the vapor with passive heat sinks.
Honestly, this was just a proof of concept. The TEC was, shall we say, rescued from one of those small fanless desktop "refrigerators". The tape was directing the air exhaust out the back. The upper fins were certainly warm, but not hot to the touch. (the added fan was doing it's job). We actually lost more fluid and wasted cooling power by the fins accidentally touching the liquid, vs. spending their energy only condensing the vapor. (The ugly masking tape wasn't exactly air tight :-)
@angleisthebest The Novec fluid is non-corrosive, and should not shorten the life of the computer. If anything, it should make it last longer, because all surfaces of the motherboard are not allowed to rise much above the boiling point of the liquid. With air, you have to worry about air flow issues, hot spots, etc. With this, anything the liquid touches is cooled reliably, but only if the temp. of that part rises above the liquid's boiling point.
@BOBSAGATification , I'm told yes, but you have to realize the following: The fluid isn't absorbing heat and carrying it away in liquid form, it phase changes to gas at a much lower temp than water. As it passes over the hot areas gas will largely be emerging from the block's other side, pressurizing the cooling system. You have to get the gas through a radiator and re-condensed, before it finds a way to escape. Most computer liquid cooling systems aren't designed to operate under pressure.
@ChillaxedJosh That's what I thought at first also, but actually, it significantly lowers the efficiency. You're thinking surface area for air... but in this case, the heat is remove directly from the CPU, and the CPU doesn't have to push the heat through a block of aluminum first.
@farmdve , Onboard video. Adding a video card would have required a lot more volume inside the tank. For people considering this with a video card, I would suggest a riser card that will allow the card to lay down over the motherboard, thus conserving space / volume. The drawback is having your video card blocking the surface of the motherboard will screw up the "aesthetics" of it. (Cooling efficiency would be just fine)
This would be a cooler test on a real processor. Most who would even think of this setup would be high-end or server usage. Lets see a high-end system keep those kind of temps. If you can keep a i7 @ 4.5ghz or so that cool @ 100% load then I would say you got something there.
1) Turn off computer (you don't want to overheat chips) 2) Remove cover, and lift board out of liquid, and shake gently 3) Count off "One Mississippi, Two Mississippi, Three Mississippi...", maybe up to 5. 4) The board is now 100% dry, and free of any evidence whatsoever that it was ever submerged in the liquid. No joke, it's really that simple and fast. :-)
Not sure of other liquids. It's pretty specialized to be non-toxic, non-conductive, non-atmospheric degrading, and have a variable boiling point depending on the chosen liquid. One thing people do, is provide a more efficient boiling surface for the chip. Not a heatsink (there is some confusion there in the thread), but a surface that promotes boiling. This can be done by adding a boiling plate, or painting chip surfaces with a chemical that promotes even boiling. (so liquid does not superheat)
@vipondiu I'd say just use a Novec with a lower boiling point. Allowing pressure, and accurately controlling pressure in the chamber will add a LOT of unnecessary complication and cost to the rig. Also I suggest using 7200 RPM laptop drives that can take higher temps, and save space. Remember this kind of cooling is different from air, the liquid temp. does not directly translate to CPU temp like air does. A warm bath of liquid is fine, and will be more tolerant of higher ambient temperatures.
This is the same way the human body cools itself... This is the same way they cool nuclear power plants... This is a great idea! Essentially, the liquid ensures that your processors can't get too hot, and if you had some way of controlling the pressure inside the vessel, you could control the temperature of the liquid itself. I know gaming enthusiasts would love to see thier GTX 480 foam and froth as it stays at 35C, but this would be even better for huge server setups.
@xG33Kx , the temps would actually be higher. You're thinking air cooling, this is different. By adding a heat sink, it acts as an insulator, making the CPU have to push the heat through a less efficient surface first, before being carried away. By allowing the liquid to contact the CPU surface,the heat is immediately carried away from as close to the point that it is generated as possible. A heat sink does work (even with a fan), but it doesn't work as well.
@CorvalentCorp I'm not thinking of water, I understand how the fluid cools. I was most concerned that if a critical temperature was reached then a gas layer might form, which will have significantly less heat transfer ability, though based on what you just said, that isn't an issue on this scale. Thx for the reply anyway.
@TheCallMeCrazy, kind of.. but no compressor. Temp changes are enough to easily change the phase of this stuff. Efficiency comes from larger scale systems... Think, 100s of blade servers in a bath of this stuff making bubbles (in a container like a deep freezer) Radiator sits over the bath, with a fluid pumped through, to outside air, where a fan cools another radiator. You are now cooling 1000s of watts server heat, with no compressors, no huge HVAC systems, no dirty fans, no mess, no noise.
@Zachstar2000, I don't know why it's so much, I suppose because it probably takes very specialized, expensive equipment to produce the chemical. They do make it in large volumes, almost the exact same chemical (under a different division) is used in place of Halon gas for server room / high value, fire suppression systems. Stays liquid until pumped through nozzles, which change it to gas, and put out amazingly fast. Apparently much safer to be exposed to vs. Halon.
Not sure how it would react... but I don't think it would produce the results you are looking for. (Bubbles, then bubbles disappear in the oil as they cool). The Novec fluid is used as a degreaser in industrial applications, and will leech the oils out of the plastics used for wiring insulation. (does not harm the wires/insulation, the oil is a byproduct of manufacturing). So I'm thinking, it would probably completely change the characteristics of the whole thing, and waste the fluid.
Do you suppose putting the TEC directly in the fluid would help? As a chemistry student, I know that the vapor would naturally condense back when the fluid temperature drops, and it allows for the possibility of lower fluid temperatures near the processor, possible alleviating the "early bubbling" problem.
It's not about how cold it is. The maximum allowable temperature is determined by the boiling point of the liquid. (plus pressure, etc. but we won't get into that). As long as the temperature is withing the operating range of the CPU, it's good, because it can't get any hotter. (as long as it's boiling) There is a reason that blacksmiths don't put their red hot iron work in front of a fan, but instead dunk it in water. Rapid liquid boiling removes heat at a *much* faster rate vs. air cooling.
I'm a bit fuzzy on these details. I believe the plate is sort of 'baked' on, as part of the chip manufacturing process, or, if retrofitting, a paint concoction can be made that has (I think) micro polymer beads in it, that produces a surface that promotes even boiling. Otherwise you end up with superheated areas of the chip that don't boil, while other parts do. One of the first things I asked about was roughing up the surface, and was told it would help, but there were better methods.
@Twisted86 Considering that 200cc of fluid can cool 4000watts of power, (according to 3M's testing), it could do it. It's just a matter of being able to condense the vapor, thus removing the heat effectively.
@TheShoura, Ah I see. I'm not sure if it would work or not. Usually in those situations you wouldn't want it to boil, just carry heat away... (like car fluid systems don't boil unless losing pressure) the tricky part would be ensuring that the liquid re-condenses effectively inside the radiator. Plus, increasing pressure, increases the boiling point in a closed system. If it's an open vented system, you lose fluid. It might be tricky, but certainly would be interesting to test.
@CorvalentCorp I get it cause I understand the properties of novec ...I work in a magnesium die cast facility where we use novec gas which uses the same properties not to cool but to mask the molten metal surface and the gas is the answer ...You probably should mention that novec is also a controlled substance and very pricey to those with the Idea of building a similar outfit.
@CorvalentCorp Well, now that I imagine, this is just an intermarriage between two exotic attempts for watercooling: TEC cooling and liquid submersion. The dripping effect with the Peltier cooler is brilliant, though. To be fair, people spend well over $700 on an exotic cooling solution like water cooling, or even up to $2000 for phase change. I think it might be a bit more realistic than expected.
@ricco197332 The flavors of Novec we were using aren't controlled, and are considered non-toxic and harmless to the environment. Indeed it is expensive though... however for a unique industrial chemical produced in relatively low volumes probably not too insane.
If you connected a garden hose to a fire hose, it would not make the water flow any faster. All the heat still has to pass through the limited surface area of the CPU. With air, it makes sense, to spread the heat out so it can be blown away. With this technology, it is removing the heat as fast as the liquid flows over the metal, closer is better. Novec is much more effective at removing heat than aluminum, so the liquid touches it first, and we avoid the bottle neck of CPU grease and aluminum.
@bozzza69 , The temps would be higher with a heat sink. The boiling is what does it, and the closer you can boil to the source of the heat, the more effectively you can remove it. The comments go into much greater detail about why this is so, further down... see if you can organize them by 'thread' (it makes more sense that way)
@cocks308 , I don't remember which version of the fluid it is closest to, but search UA-cam for: "Novec 1230" or "3M Fire Fluid" and you will find a lot of results for that application.
Bravo! Absolutely amazing demonstration! Also a big thumbs up to the video producer for great choice of music and cut in effects showing how amazing this is. A quick question one of the worries about these total immersion systems is metal leaching from various parts and contaminating the liquid which will eventually short out the board. Have you noticed any of that or run any tests on contamination over time?
If you sort the comments by thread, it's covered quite a bit in the previous comments. Short answer: It could handle it... but for best results one should treat the surface of the chips to promote even boiling. (boiling plate). More boiling = more cooling.
@haghighihk Yes, a large passive heat sink would work. You would probably want to use a version of the fluid with a slightly higher boiling point to make sure the vapor / heat has a greater difference than the ambient temperature. I don't see the cost of this particular fluid changing very much. It is a highly specialized, industrial chemical from 3M. There are specialized applications where it makes a lot of sense, but for your average everyday consumer desktop PC, it's not very cost effective.
@LechuCzechu Yup. Highly impractical for any kind of average consumer use. But interesting none-the-less. Huge efficiency gains are possible in a very dense server type environment, by eliminating fans and facility HVAC systems, moving the heat directly from computers to outside air. (see comment history below)
@MADJIHAD1942 This test wasn't ever about efficiency. Especially considering the amperage used by the TE. Need efficiency? 1) Switch to a higher boiling point Novec: 649(49C) or 7100(61C) 2) Use standard liquid radiator cooling to condense the vapor, and carry heat to outside air 3) 100's of tightly packed blade servers in a bath of Novec, in a container design like a deep freeze 4) ??? 5) Profit: You're cooling 1000's of watts with No dirty wasteful fans, huge HVAC systems, or Noise
Expensive~$300 gal. Not sure what you're asking, there is no water is in this test / video. You could use a standard water/radiator+fan loop in place of the TE cooler (3M made this work successfully), but we have not tested it. The idea is only to condense the vapor, not to chill the liquid. You *want* the liquid to boil, that's how the heat is carried away. If the TE cooler fins / radiator to touch the liquid it really hurts efficiency (we lost some of the fluid this way before we caught on)
@rokr41 , We didn't try. In theory, one should be able to push it until you get so hot that the fluid vaporizes before touching the chip, then your temps would begin to rise faster than that the fluid can flow in to cool. Maybe active pumping of the fluid "squirting" it at the chip, might overcome that (guessing). As a company, we're more interested in long-term production, dependability & long-term reliability (particularly in extreme environmental conditions), vs. huge overclock speeds :-)
How does it cool with an OEM Heat Sink (rather than some aftermarket) HS? I would think with a basic HS it would get good flow, and still reach the Boiling point to allow for cooling?
To more directly answer your question, it will cool and still works, but not nearly as well because you are limiting the performance of the liquid by forcing the heat to pass through less effective/efficient heat conductors first. There are paints, and surface treatments that can be done to the CPU to encourage better, more even boiling. This further increases the liquid's effectiveness, because: bubbles = cooling. No bubbles, no cooling. (meaning the surface is below the liquid boiling point)
We could potentially use this stuff to replace the current fluid in heat pipes. And instead of using capillary action it could just boil up to a collector/radiator, right?
I'm not sure. I've not heard this. I would imagine that if this were the case, the rate of evaporation would be extremely slow. It's much more likely one will lose fluid through expansion and contraction of the gas as the system pressurizes due to thermal changes. A bellows of some type would be important, unless your temp control is very very good.
@CorvalentCorp Thanks for the answer! Pertaining to the possibilities, I wonder if this is something that will possibly eventually be used for workstation type computers? It would definitely be a neat experiment and would be great as a long term cooling solution.
@dcyli , Possibly, you would have to be very careful how it was oriented. I'm not sure how well it would work in such a constricted system without some form of active pumping. Another method of using this is to spray the liquid on the board (some Cray supercomputers do this, as do some MIL aircraft computers). It uses much less expensive fluid, but I don't like being dependent on pumps. (And should your pump fail, when the chips have no other form of heat dissipation, it could be very bad.)
@CorvalentCorp You're right about HDDs, they're not fully sealed, so I'll have to figure out how to cool them (right now they are near 50C, the limit for recomended operation) What about using Novec 649 at...something like 0,7 atm so the boiling point gets down to 35-40C aprox? do you think it's feaseable?
@backstyck Some supercomputers use a spray method. I like submersion, because it temperature limits all parts of the board, nothing can exceed temp X unless it's insulated by something like a big heat sink, and a layer of grease. :-) 3M successfully cooled 4000w of heat using 200cc of the liquid. So there is enormous potential. We just did a proof of concept. Remember the heat still has to be removed and the vapor condensed, or you lose fluid through the gas expansion pressure.
@backstyck It's exotic vs. air cooling, needs to be in a sealed container, with IO sockets out of the housing and the heat still has to be processed out (the vapor condensed) It's a more complex rig & to be efficient, needs redesigned Motherboards. Imagine 100s of super thin flat computers packed into a bath of it. A complete data center of computing packed into a file cabinet size space, w/standard water radiator cooling to outside air. No Fans, No big HVAC, No noise. Huge savings.
seeing as this will only be used by overclockers for the sake of it in the home market, would a heat sink not help if the chip was massively clocked, as the chip can only boil away so much before the boiling of gas gets in the way, a boiling plate as you say would help, but a standard heat sink would give you so much surface area that you could remove more heat... although it would never reach as cold, it would stop it from getting hotter as there is space for more boiling... would it not
Are there any alternatives to TECs for cooling (well obviously) say for example a compressor? that might be more efficient. Also, for a CPU say staying below 70oC what is the maximum Wattage that the fluid would be able to deal with for overclocking purposes? I'm just thinking an OC'd i7 can get through over 200W. So to be safe you'd need at least 250W of cooling potential on the area of the CPU.
@Hokrollo1 (I thought that at first too) In reality, the heat has to be pushed through the heat sink, which is less efficient than the liquid at cooling. It will still cool, but with less efficiency. A bad analogy, is that it would be like a blacksmith trying to cool a piece of red hot iron by first inserting it in an aluminum or copper sleeve, and only then dunking it in cold water. The straight liquid is much more efficient as long as it has a good surface on which the fluid can boil.
@rokr41 It could be, but it would not be practical at this point. This is about achieving desktop computing power in very extreme environments (dusty, high pressure deep underwater, very hot, very humid, etc.) and it's also about super efficiency if used to cool hundreds of server boards packed tightly together in the liquid. For a single desktop, office type system, it's cool to look at, and fun to make, and cools very effectively, but just not practical or efficient from a cost standpoint.
oh darn it.... 300 a gal? I actually just build a oil submerged system with 4 giant Thermoelectric cooling plus crosair H100 water cooling system laying flat down and completely air tight. The setup is PERFECT for this novec thing. I was thinking maybe this novec liquid is lighter than oil, if i could cool the oil enough, my novec would recondence in the oil and not have a chance to escape my pleix glass or any other gaps.
@TheRussianCinema Would the price not come down tho with major orders? I would HIGHLY doubt it would be 300USD a gal at 1000 gal or more. Right now it sounds like this stuff is prepared basically by hand as opposed to highly automated chemical manufacturing.
So do you have pictures of what this boiler plate will look like over the CPU? How will a boilerplate be better than a typical heatsink? Have you attempted to overclock it? How high can you overclock this processor in this fluid before you start experiencing problems?
English aint my native language but anyway: Q1: How is the thermal conductivity of 3M's liquid compared to air, water & baby oil? Q2: Will one be able to pick up the components and use them in a normalcooling way again? Wondering since similar solutions with baby oil is more common but has the negative consequence that it sticks to the parts and are difficult to remove, making it difficult to use parts in the "normal" way again once you have submuged them ones.
That would probably work quite well. Some kind of bellows, balloon, etc. Although I'm not sure how the chemical reacts with latex, it might need to be some other material.
Not sure if this has been asked before but how much loss of fluid do you get? or is this a sealed system? I would think you would get a bit of fluid loss with all the cables being routed to the top of the container for the HDD. I actually had a very similar idea using the same fluid from 3M, I was insure about how to cool the gas I was thinking of a type of expansion chamber/condenser, but the peltier is a very good idea.
@miest0908 Bubbles = cooling. No bubbles, then temp is not above boiling. If the TE is in the liquid, it then sucks the cooling power to cool EVERYTHING. Hotspots still continue making bubbles which are then not condensed quickly enough because the cooling power is going to the liquid vs vapor. Pressure rises, temp rises, & bad things happen (air leak/rupture, loss of fluid, loss of temp control). In our test, the TE was touching the liquid, an we lost a lot of fluid before we figured it out.
Those bubbles are really bad for cooling. The vapor acts as an insulator for the components, trapping the heat in. My father worked with supercomputers and helped design the method of liquid cooling most supercomputers employ; instead of submersing the board, put it at an angle (30 degrees works well, IO at the bottom, but it shouldn't matter too much), and allow the liquid to cascade, aided by a pump. You should decrease you temps by at least 10 degrees.
One Mississippi, Two Mississippi, Three Mississippi... Yes. The computer runs fully submerged in the liquid, and it is non-conductive. Although it is clear, it is not water. So being semi-"wet" would make no difference. The concern would be the near immediate overheating of the chips once removed, hence, turning it off first.
Hi, I'm thinking of doing that to my PC (not an OC fan, but extremely tired of removing dust in every tiny corner of heat sinks). I think of using a dielectric with higher-boiling point but reduce the pressure inside the box (cause it has to be hermetic anyways) so it boils at a custom temperature. By the way, I think HDDs can be submerged since thay are hermetic, i've seen that in oil-submerged pcs, haven't dare myself.
what about cavitation from the air bubbles. when air bubbles constantly explode up next to an item causing cavities. -------great product really awesome idea.
I would not suggest it as a coolant, wile it is much cheaper than dielectric coolants. Novec's are rather expensive HFE solvents, they will will slowly eat into the layer binding and solder mask. GM along with a number of other companies do make a dielectric coolant designed for submersion cooling servers. It runs about 1,000USD a gallon.
So technically its the phase change of this fluid boiling off the processor that keeps it cool, but also I hear that this chemical transfer heat without changing phases fairly well. Maybe even better than mineral oil. Have you tried running the processor with the fan and heatsink installed as normal? What results do you get with graphics cards?
Also 1 more question... How much power are you dissipating with this set up? You have inspired me a bit to actually try this out now, and I would like to see how well it would hold up to cooling as you increase power. Even if you could just give me the voltage and architecture your running in this build. Lastly how much current are you putting through your peltier to cool the gas enough to condense it? I would be interesting to see if this is a bit of a power saver =)
@kyle2194 , It actually makes it worse to have a heat sink. With the boiling liquid, you are immediately sapping the heat directly from the chip as it boils. If you put a heat sink on it, then you lose this ability, and the chip has to heat *through* the heat sink first. (actually has an insulating effect) It will still cool, just not quite as effectively. Weak analogy: If you jump into freezing water with an aluminum suit on Vs. bare skin, which would chill your body down most effectively?
That is actually really cool :) I'm wondering a few things: How hot were the CPU VRMs? Would mounting a large, flat piece of copper help? Will you be repeating this test with higher TDP processors?(115w+)
@malvcalv , Heh, heh, no we didn't drink it. I can think of a lot of things that are not toxic, that I still wouldn't want to drink. :-) I've put my hands in it, smelled it, etc. (virtually no smell), and it evaporates off your hands almost immediately. A version of this almost exact same chemical is used in fire suppression systems for server rooms in place of Halon gas. There are some interesting videos of it putting out fires.
I know the comments are old, but I think what he meant is, that when there is extreme boiling, the bubbles of air will use too much espace "arround" the chip surface, and that air would prevent the liquid to make contact with the chip surface. That would in theory reduce the efficiency.... While in the other hand, a bigger heatsink (well seated on the chip) would give a bigger surface for the liquid to make contact and to dissipate the heat.
I will agree somewhat on the "no practical cooling advantages". For a desktop system, it is not cost effective, and completely not practical compared to air cooling. It's just interesting. If you want practical, read down in the thread about computer data centers.
Imagine 100 blade PCs lined up in a Novec bath, spaced a few mm's apart. No heatsinks, no energy wasted on fans. Vapor condensed by conventional water radiator, that carries heat outside. Now, no HVAC systems are needed. BIG SAVINGS
@kristofferD
1) It's not about transferring the heat to the liquid (as is the case with oil cooling, or water cooling) With this, it is about the phase change from liquid to gas that transfers the heat. The liquid has a very low boil point, so it cools only the areas (bubbles) that have greater heat than the liquid's boiling point.
2) No the liquid evaporates in seconds, and the board is 100% dry to the touch, nothing remains behind, no residue, no oils, no dust, no rust, nothing
@TheScyy, we did not test video cards... but there is no reason it would not work. Tests have been done at 3M, where only 200cc of this fluid (the version with a 49C boiling point) was able to continuously cool over 4000watts. ( +/- equivalent of (30) i7940 CPUs @ 100% load).
Remember this isn't magic, you still have to do something with all the heat. (increase the capacity of the TE coolers on the top, or switch to a liquid cooled radiator to move the heat out and condense the vapor.)
@backstyck : I suppose it could get so hot that it produces a layer of gas around the chip, but the expanding gas leaves (rises) and is replaced, so cooling still occurs even if the fluid does not directly contact the chip. It's the phase change, the actual act of boiling, *not* direct heat transfer to the liquid fluid, that makes this particular process so effective. Cooling the liquid is a waste of energy, you need to condense the vapor to complete the phase change process.
@Uber1nsane, I'll see if I can find out and post it. We're an industrial computer system manufacturer, and got it directly from 3M. Depending on the "flavor" ~$300 per gallon give or take. That's why we used the foam (it was to displace enough liquid so that 1 gallon would be enough)
The fluid can dissipate over 4000 watts of heat with only 200cc of fluid. People serious about using this stuff, have to design computer hardware to be very dense, then it's crazy efficient compared to fans.
@thebeetalls : Not sure. We were watching the temp. of the liquid, & the temp that the CPU was reporting. By the nature of the fluid, it boils at a specific temperature. It's the boiling that removes heat, provided the fluid is replaced (because it's submerged) this limits the "allowed" heat of any part on the board.
Treating the CPU surface with a chemical boiling plate, will greatly improve the cooling delta. It's not about increasing the surface area, but rather ensuring boiling across it.
@CaptainDishman We didn't overclock. However in theory, the temps should remain constant. The liquid always boils at the same temp, and should maintain low temps indefinitely provided the vapor is being condensed quickly enough that you don't run out of fluid.
@PortaleMedia , It was running "BurnInTest v6.0" You're right though, "Speed Fan" was installed. Can't easily run the tests using other software, we already disassembled that rig. Hopefully I'll be building a version soon, that is *much* more interesting than the proof of concept.
We monitored the various temperatures, as well as the fluid temperature (the multimeter in the video, showing Fahrenheit temp. from a probe) the whole time. It's was all good.
@crabid yes, see lower thread comments:TEC is compact but inefficient. Other methods are possible, even 100% passive w/enough heat sink surface area. You just have to get the heat out somehow (condensing the vapor)
2nd: You're thinking water. here, boiling = cooling. CPU is always 5-18C over boiling (varies depending on CPU surface treatment used to encourage efficient boiling) It can't get hotter. 3M did a successful test cooling 4000w, w/only 200ml of fluid. (same as ~30 i7940's @ 100% load)
@LechuCzechu Sort of. Allowing the container to pressurize and then trying to regulate that pressure would greatly increase the complexity. Venting the pressure = fluid loss. Fluid = ~$300gal. The fun thing with this is that is temp is regulated by boiling point, and the whole board, all parts have equal cooling.
@vipondiu, There are different types of Novec fluid, some with higher boiling points. Novec 649 boils at 49C, and will be what we use if we revisit this project. With a good boiling surface on the CPU, the chip core temp will be somewhere around 55 to 62C.
Hard drives are sealed from dust, but not atmospheric pressure. They all have vent holes to equalize the pressure. Those can be sealed, and the hard drives submerged, but no guarantees on disk life, performance or warranty after that. :-)
@crabid No problem :-) Even with a gas layer, consider how quickly red hot metal cools in water, just a few seconds for the boiling to stop (~600C to >100C in almost an instant) then consider how much longer the metal remains hot to touch (when not boiling, the rate of cooling is much slower)
Now consider that this liquid (depending on the flavor) boils at ~1/2 to well under 1/2 the temp of water. That means this very aggressive cooling is happening at, and maintaining, much lower temps.
@otoluk In our test, the CPU was not treated to encourage even, complete boiling, so the numbers are not as good. There is a special paint that can be applied to the surface, or, a boiling plate can be bonded to the chip. (Not to add surface area, but to provide a good boiling surface)
If memory serves, when that is done the temp of the CPU compared to the temp of the liquid boiling point changes to somewhere between 2 and 5 degrees over the fluid boiling temp. (depending on treatment method)
Followup: Talked to 3M, and found out that the oils in the flexible PVC plastics are needed only during manufacturing, and do not really affect the performance of the plastic. So while it leeches the oils out of the plastics over the long-term, it does not cause them to fail or become brittle. At the worst, you'll end up with some oil droplets condensing on the walls. (placing activated carbon (a fishtank filter) in the liquid, will absorb the waste oil if this is a problem)
yes. If you click the "sort by thread (beta)" link, it will organize these comments into a conversation that can be read where a lot of questions are answered. Cost wise, if you are not phase changing, some other substance would be better that doesn't evaporate or cost so much. We did not test with graphics cards or overclocking. Just a proof of concept. :-)
Boiling is a method of cooling or maintaining temperatures. If a liquid will boil at a given temperature, exceeding that temperature converts the liquid to a gas. When you pour water on a hot pan, and the water turns to steam, it cools the pan... very rapidly.
Similarly, this liquid boils at a very low temperature, so it will turn to vapor much more quickly, thus removing the heat from the hot surfaces. The difference is, this liquid does not conduct electricity, so it is harmless to a PC.
Our engineering department is a bit high on ambient noise. (various fans, beeping, etc.) But generally speaking, it's dead silent, particularly with the cover on it. All components could be submerged, but one has to deactivate speed sensors. The heat still has to be processed out, so there may be noise depending on your chosen cooling method. Unless you use something like a remote water pump to cool the condenser heat sink, or a larger passive heatsink (with higher boiling point liquid).
@backstyck 3M cooled 4000watts of heat in an 8x8" area, w/200ml of fluid. (~equivalent of 30, i7940's) Remember, it's not heat to liquid (like in mineral oil cooling). You're vaporizing the liquid to move the heat, it's the phase change that does it. I guess jets might help if it's just crazy hot, but a larger boiling plate (not a heat sink) would probably help more. The idea is to instantly boil the heat away before it get's hot enough to create the massive cavitation bubbles.
@mranenome, some of the sealed "Heat Pipe" technology uses a similar principal.
The "magic" here, comes from the phase change. Most conventional computer liquid cooling systems are designed for the liquid to remain in liquid form, and would not deal well with pressure from the expanding gas.
Exactly. It limits the temperature of any part of the board... it's like a giant super-effective heat sink covering every component from every side. The temperature delta depends on how effectively the surface of the component allows boiling. If we do it again, we'll probably go with a fluid with a slightly higher boiling point, so we can more readily condense the vapor with passive heat sinks.
Honestly, this was just a proof of concept. The TEC was, shall we say, rescued from one of those small fanless desktop "refrigerators". The tape was directing the air exhaust out the back. The upper fins were certainly warm, but not hot to the touch. (the added fan was doing it's job).
We actually lost more fluid and wasted cooling power by the fins accidentally touching the liquid, vs. spending their energy only condensing the vapor. (The ugly masking tape wasn't exactly air tight :-)
@angleisthebest The Novec fluid is non-corrosive, and should not shorten the life of the computer. If anything, it should make it last longer, because all surfaces of the motherboard are not allowed to rise much above the boiling point of the liquid. With air, you have to worry about air flow issues, hot spots, etc. With this, anything the liquid touches is cooled reliably, but only if the temp. of that part rises above the liquid's boiling point.
@BOBSAGATification , I'm told yes, but you have to realize the following: The fluid isn't absorbing heat and carrying it away in liquid form, it phase changes to gas at a much lower temp than water. As it passes over the hot areas gas will largely be emerging from the block's other side, pressurizing the cooling system. You have to get the gas through a radiator and re-condensed, before it finds a way to escape. Most computer liquid cooling systems aren't designed to operate under pressure.
@Ryday37 No rust. It's not water, and there is nothing in the liquid to cause oxidation. So no issues there. :-)
@ChillaxedJosh That's what I thought at first also, but actually, it significantly lowers the efficiency. You're thinking surface area for air... but in this case, the heat is remove directly from the CPU, and the CPU doesn't have to push the heat through a block of aluminum first.
@farmdve , Onboard video. Adding a video card would have required a lot more volume inside the tank. For people considering this with a video card, I would suggest a riser card that will allow the card to lay down over the motherboard, thus conserving space / volume. The drawback is having your video card blocking the surface of the motherboard will screw up the "aesthetics" of it. (Cooling efficiency would be just fine)
This would be a cooler test on a real processor. Most who would even think of this setup would be high-end or server usage. Lets see a high-end system keep those kind of temps.
If you can keep a i7 @ 4.5ghz or so that cool @ 100% load then I would say you got something there.
@drdrdrc , good question. Not sure. We didn't see any such issue, nor has 3M as far as I know.
1) Turn off computer (you don't want to overheat chips)
2) Remove cover, and lift board out of liquid, and shake gently
3) Count off "One Mississippi, Two Mississippi, Three Mississippi...", maybe up to 5.
4) The board is now 100% dry, and free of any evidence whatsoever that it was ever submerged in the liquid.
No joke, it's really that simple and fast. :-)
Not sure of other liquids. It's pretty specialized to be non-toxic, non-conductive, non-atmospheric degrading, and have a variable boiling point depending on the chosen liquid.
One thing people do, is provide a more efficient boiling surface for the chip. Not a heatsink (there is some confusion there in the thread), but a surface that promotes boiling. This can be done by adding a boiling plate, or painting chip surfaces with a chemical that promotes even boiling. (so liquid does not superheat)
@vipondiu I'd say just use a Novec with a lower boiling point. Allowing pressure, and accurately controlling pressure in the chamber will add a LOT of unnecessary complication and cost to the rig. Also I suggest using 7200 RPM laptop drives that can take higher temps, and save space. Remember this kind of cooling is different from air, the liquid temp. does not directly translate to CPU temp like air does. A warm bath of liquid is fine, and will be more tolerant of higher ambient temperatures.
This is the same way the human body cools itself... This is the same way they cool nuclear power plants... This is a great idea! Essentially, the liquid ensures that your processors can't get too hot, and if you had some way of controlling the pressure inside the vessel, you could control the temperature of the liquid itself.
I know gaming enthusiasts would love to see thier GTX 480 foam and froth as it stays at 35C, but this would be even better for huge server setups.
@xG33Kx , the temps would actually be higher. You're thinking air cooling, this is different. By adding a heat sink, it acts as an insulator, making the CPU have to push the heat through a less efficient surface first, before being carried away. By allowing the liquid to contact the CPU surface,the heat is immediately carried away from as close to the point that it is generated as possible. A heat sink does work (even with a fan), but it doesn't work as well.
@CorvalentCorp I'm not thinking of water, I understand how the fluid cools.
I was most concerned that if a critical temperature was reached then a gas layer might form, which will have significantly less heat transfer ability, though based on what you just said, that isn't an issue on this scale.
Thx for the reply anyway.
@TheCallMeCrazy, kind of.. but no compressor. Temp changes are enough to easily change the phase of this stuff. Efficiency comes from larger scale systems... Think, 100s of blade servers in a bath of this stuff making bubbles (in a container like a deep freezer) Radiator sits over the bath, with a fluid pumped through, to outside air, where a fan cools another radiator. You are now cooling 1000s of watts server heat, with no compressors, no huge HVAC systems, no dirty fans, no mess, no noise.
@Zachstar2000, I don't know why it's so much, I suppose because it probably takes very specialized, expensive equipment to produce the chemical. They do make it in large volumes, almost the exact same chemical (under a different division) is used in place of Halon gas for server room / high value, fire suppression systems. Stays liquid until pumped through nozzles, which change it to gas, and put out amazingly fast. Apparently much safer to be exposed to vs. Halon.
Not sure how it would react... but I don't think it would produce the results you are looking for. (Bubbles, then bubbles disappear in the oil as they cool).
The Novec fluid is used as a degreaser in industrial applications, and will leech the oils out of the plastics used for wiring insulation. (does not harm the wires/insulation, the oil is a byproduct of manufacturing). So I'm thinking, it would probably completely change the characteristics of the whole thing, and waste the fluid.
Do you suppose putting the TEC directly in the fluid would help? As a chemistry student, I know that the vapor would naturally condense back when the fluid temperature drops, and it allows for the possibility of lower fluid temperatures near the processor, possible alleviating the "early bubbling" problem.
It's not about how cold it is. The maximum allowable temperature is determined by the boiling point of the liquid. (plus pressure, etc. but we won't get into that). As long as the temperature is withing the operating range of the CPU, it's good, because it can't get any hotter. (as long as it's boiling)
There is a reason that blacksmiths don't put their red hot iron work in front of a fan, but instead dunk it in water. Rapid liquid boiling removes heat at a *much* faster rate vs. air cooling.
I'm a bit fuzzy on these details. I believe the plate is sort of 'baked' on, as part of the chip manufacturing process, or, if retrofitting, a paint concoction can be made that has (I think) micro polymer beads in it, that produces a surface that promotes even boiling. Otherwise you end up with superheated areas of the chip that don't boil, while other parts do. One of the first things I asked about was roughing up the surface, and was told it would help, but there were better methods.
@Twisted86 Considering that 200cc of fluid can cool 4000watts of power, (according to 3M's testing), it could do it. It's just a matter of being able to condense the vapor, thus removing the heat effectively.
@TheShoura, Ah I see. I'm not sure if it would work or not. Usually in those situations you wouldn't want it to boil, just carry heat away... (like car fluid systems don't boil unless losing pressure) the tricky part would be ensuring that the liquid re-condenses effectively inside the radiator. Plus, increasing pressure, increases the boiling point in a closed system. If it's an open vented system, you lose fluid. It might be tricky, but certainly would be interesting to test.
@CorvalentCorp
I get it cause I understand the properties of novec ...I work in a magnesium die cast facility where we use novec gas which uses the same properties not to cool but to mask the molten metal surface and the gas is the answer ...You probably should mention that novec is also a controlled substance and very pricey to those with the Idea of building a similar outfit.
@CorvalentCorp Well, now that I imagine, this is just an intermarriage between two exotic attempts for watercooling: TEC cooling and liquid submersion. The dripping effect with the Peltier cooler is brilliant, though.
To be fair, people spend well over $700 on an exotic cooling solution like water cooling, or even up to $2000 for phase change. I think it might be a bit more realistic than expected.
@ricco197332 The flavors of Novec we were using aren't controlled, and are considered non-toxic and harmless to the environment. Indeed it is expensive though... however for a unique industrial chemical produced in relatively low volumes probably not too insane.
If you connected a garden hose to a fire hose, it would not make the water flow any faster. All the heat still has to pass through the limited surface area of the CPU. With air, it makes sense, to spread the heat out so it can be blown away. With this technology, it is removing the heat as fast as the liquid flows over the metal, closer is better. Novec is much more effective at removing heat than aluminum, so the liquid touches it first, and we avoid the bottle neck of CPU grease and aluminum.
@bozzza69 , The temps would be higher with a heat sink. The boiling is what does it, and the closer you can boil to the source of the heat, the more effectively you can remove it. The comments go into much greater detail about why this is so, further down... see if you can organize them by 'thread' (it makes more sense that way)
@cocks308 , I don't remember which version of the fluid it is closest to, but search UA-cam for: "Novec 1230" or "3M Fire Fluid" and you will find a lot of results for that application.
Bravo! Absolutely amazing demonstration! Also a big thumbs up to the video producer for great choice of music and cut in effects showing how amazing this is.
A quick question one of the worries about these total immersion systems is metal leaching from various parts and contaminating the liquid which will eventually short out the board. Have you noticed any of that or run any tests on contamination over time?
If you sort the comments by thread, it's covered quite a bit in the previous comments. Short answer: It could handle it... but for best results one should treat the surface of the chips to promote even boiling. (boiling plate). More boiling = more cooling.
@haghighihk Yes, a large passive heat sink would work. You would probably want to use a version of the fluid with a slightly higher boiling point to make sure the vapor / heat has a greater difference than the ambient temperature. I don't see the cost of this particular fluid changing very much. It is a highly specialized, industrial chemical from 3M. There are specialized applications where it makes a lot of sense, but for your average everyday consumer desktop PC, it's not very cost effective.
@LechuCzechu Yup. Highly impractical for any kind of average consumer use. But interesting none-the-less. Huge efficiency gains are possible in a very dense server type environment, by eliminating fans and facility HVAC systems, moving the heat directly from computers to outside air. (see comment history below)
@MADJIHAD1942 This test wasn't ever about efficiency. Especially considering the amperage used by the TE. Need efficiency?
1) Switch to a higher boiling point Novec: 649(49C) or 7100(61C)
2) Use standard liquid radiator cooling to condense the vapor, and carry heat to outside air
3) 100's of tightly packed blade servers in a bath of Novec, in a container design like a deep freeze
4) ???
5) Profit: You're cooling 1000's of watts with No dirty wasteful fans, huge HVAC systems, or Noise
Expensive~$300 gal. Not sure what you're asking, there is no water is in this test / video. You could use a standard water/radiator+fan loop in place of the TE cooler (3M made this work successfully), but we have not tested it. The idea is only to condense the vapor, not to chill the liquid. You *want* the liquid to boil, that's how the heat is carried away. If the TE cooler fins / radiator to touch the liquid it really hurts efficiency (we lost some of the fluid this way before we caught on)
@rokr41 , We didn't try. In theory, one should be able to push it until you get so hot that the fluid vaporizes before touching the chip, then your temps would begin to rise faster than that the fluid can flow in to cool. Maybe active pumping of the fluid "squirting" it at the chip, might overcome that (guessing). As a company, we're more interested in long-term production, dependability & long-term reliability (particularly in extreme environmental conditions), vs. huge overclock speeds :-)
How does it cool with an OEM Heat Sink (rather than some aftermarket) HS? I would think with a basic HS it would get good flow, and still reach the Boiling point to allow for cooling?
To more directly answer your question, it will cool and still works, but not nearly as well because you are limiting the performance of the liquid by forcing the heat to pass through less effective/efficient heat conductors first.
There are paints, and surface treatments that can be done to the CPU to encourage better, more even boiling. This further increases the liquid's effectiveness, because: bubbles = cooling. No bubbles, no cooling. (meaning the surface is below the liquid boiling point)
We could potentially use this stuff to replace the current fluid in heat pipes. And instead of using capillary action it could just boil up to a collector/radiator, right?
I'm not sure. I've not heard this. I would imagine that if this were the case, the rate of evaporation would be extremely slow. It's much more likely one will lose fluid through expansion and contraction of the gas as the system pressurizes due to thermal changes. A bellows of some type would be important, unless your temp control is very very good.
@CorvalentCorp Thanks for the answer! Pertaining to the possibilities, I wonder if this is something that will possibly eventually be used for workstation type computers? It would definitely be a neat experiment and would be great as a long term cooling solution.
@dcyli , Possibly, you would have to be very careful how it was oriented. I'm not sure how well it would work in such a constricted system without some form of active pumping. Another method of using this is to spray the liquid on the board (some Cray supercomputers do this, as do some MIL aircraft computers). It uses much less expensive fluid, but I don't like being dependent on pumps. (And should your pump fail, when the chips have no other form of heat dissipation, it could be very bad.)
@CorvalentCorp You're right about HDDs, they're not fully sealed, so I'll have to figure out how to cool them (right now they are near 50C, the limit for recomended operation)
What about using Novec 649 at...something like 0,7 atm so the boiling point gets down to 35-40C aprox? do you think it's feaseable?
@backstyck Some supercomputers use a spray method.
I like submersion, because it temperature limits all parts of the board, nothing can exceed temp X unless it's insulated by something like a big heat sink, and a layer of grease. :-)
3M successfully cooled 4000w of heat using 200cc of the liquid. So there is enormous potential. We just did a proof of concept.
Remember the heat still has to be removed and the vapor condensed, or you lose fluid through the gas expansion pressure.
@backstyck It's exotic vs. air cooling, needs to be in a sealed container, with IO sockets out of the housing and the heat still has to be processed out (the vapor condensed) It's a more complex rig & to be efficient, needs redesigned Motherboards.
Imagine 100s of super thin flat computers packed into a bath of it. A complete data center of computing packed into a file cabinet size space, w/standard water radiator cooling to outside air. No Fans, No big HVAC, No noise. Huge savings.
Pretty cool to see that thing bubbling like that!
seeing as this will only be used by overclockers for the sake of it in the home market, would a heat sink not help if the chip was massively clocked, as the chip can only boil away so much before the boiling of gas gets in the way, a boiling plate as you say would help, but a standard heat sink would give you so much surface area that you could remove more heat... although it would never reach as cold, it would stop it from getting hotter as there is space for more boiling... would it not
Are there any alternatives to TECs for cooling (well obviously) say for example a compressor? that might be more efficient.
Also, for a CPU say staying below 70oC what is the maximum Wattage that the fluid would be able to deal with for overclocking purposes? I'm just thinking an OC'd i7 can get through over 200W. So to be safe you'd need at least 250W of cooling potential on the area of the CPU.
I read that its not ideal to use plexiglass and epoxy to make your enclosure because the coolant can evaporate through the epoxy ... is that true?
@Hokrollo1 (I thought that at first too) In reality, the heat has to be pushed through the heat sink, which is less efficient than the liquid at cooling. It will still cool, but with less efficiency.
A bad analogy, is that it would be like a blacksmith trying to cool a piece of red hot iron by first inserting it in an aluminum or copper sleeve, and only then dunking it in cold water. The straight liquid is much more efficient as long as it has a good surface on which the fluid can boil.
@CorvalentCorp
Wouldn't that effectively be a phase change system, similar to an A/C unit?
@rokr41 It could be, but it would not be practical at this point. This is about achieving desktop computing power in very extreme environments (dusty, high pressure deep underwater, very hot, very humid, etc.) and it's also about super efficiency if used to cool hundreds of server boards packed tightly together in the liquid.
For a single desktop, office type system, it's cool to look at, and fun to make, and cools very effectively, but just not practical or efficient from a cost standpoint.
oh darn it.... 300 a gal? I actually just build a oil submerged system with 4 giant Thermoelectric cooling plus crosair H100 water cooling system laying flat down and completely air tight. The setup is PERFECT for this novec thing. I was thinking maybe this novec liquid is lighter than oil, if i could cool the oil enough, my novec would recondence in the oil and not have a chance to escape my pleix glass or any other gaps.
@TheRussianCinema
Would the price not come down tho with major orders? I would HIGHLY doubt it would be 300USD a gal at 1000 gal or more. Right now it sounds like this stuff is prepared basically by hand as opposed to highly automated chemical manufacturing.
So do you have pictures of what this boiler plate will look like over the CPU? How will a boilerplate be better than a typical heatsink? Have you attempted to overclock it? How high can you overclock this processor in this fluid before you start experiencing problems?
English aint my native language but anyway:
Q1: How is the thermal conductivity of 3M's liquid compared to air, water & baby oil?
Q2: Will one be able to pick up the components and use them in a normalcooling way again? Wondering since similar solutions with baby oil is more common but has the negative consequence that it sticks to the parts and are difficult to remove, making it difficult to use parts in the "normal" way again once you have submuged them ones.
That would probably work quite well. Some kind of bellows, balloon, etc. Although I'm not sure how the chemical reacts with latex, it might need to be some other material.
Not sure if this has been asked before but how much loss of fluid do you get? or is this a sealed system? I would think you would get a bit of fluid loss with all the cables being routed to the top of the container for the HDD.
I actually had a very similar idea using the same fluid from 3M, I was insure about how to cool the gas I was thinking of a type of expansion chamber/condenser, but the peltier is a very good idea.
Would adding a heatsink to increase the surface area increase the cooling performance of this liquid?
@miest0908 Bubbles = cooling. No bubbles, then temp is not above boiling. If the TE is in the liquid, it then sucks the cooling power to cool EVERYTHING. Hotspots still continue making bubbles which are then not condensed quickly enough because the cooling power is going to the liquid vs vapor. Pressure rises, temp rises, & bad things happen (air leak/rupture, loss of fluid, loss of temp control). In our test, the TE was touching the liquid, an we lost a lot of fluid before we figured it out.
Those bubbles are really bad for cooling. The vapor acts as an insulator for the components, trapping the heat in. My father worked with supercomputers and helped design the method of liquid cooling most supercomputers employ; instead of submersing the board, put it at an angle (30 degrees works well, IO at the bottom, but it shouldn't matter too much), and allow the liquid to cascade, aided by a pump. You should decrease you temps by at least 10 degrees.
One Mississippi, Two Mississippi, Three Mississippi...
Yes. The computer runs fully submerged in the liquid, and it is non-conductive. Although it is clear, it is not water. So being semi-"wet" would make no difference. The concern would be the near immediate overheating of the chips once removed, hence, turning it off first.
Hi, I'm thinking of doing that to my PC (not an OC fan, but extremely tired of removing dust in every tiny corner of heat sinks). I think of using a dielectric with higher-boiling point but reduce the pressure inside the box (cause it has to be hermetic anyways) so it boils at a custom temperature.
By the way, I think HDDs can be submerged since thay are hermetic, i've seen that in oil-submerged pcs, haven't dare myself.
@Uber1nsane
You can see our G5 Immersion Experiment too. Search "G5 immersion"
what about cavitation from the air bubbles. when air bubbles constantly explode up next to an item causing cavities. -------great product really awesome idea.
I would not suggest it as a coolant, wile it is much cheaper than dielectric coolants. Novec's are rather expensive HFE solvents, they will will slowly eat into the layer binding and solder mask. GM along with a number of other companies do make a dielectric coolant designed for submersion cooling servers. It runs about 1,000USD a gallon.
So technically its the phase change of this fluid boiling off the processor that keeps it cool, but also I hear that this chemical transfer heat without changing phases fairly well. Maybe even better than mineral oil. Have you tried running the processor with the fan and heatsink installed as normal? What results do you get with graphics cards?
This is amazing. Soon enough, with a passive Peltier cooler, and a price drop, we could have an entirely passive computer submerged in water.
Also 1 more question...
How much power are you dissipating with this set up? You have inspired me a bit to actually try this out now, and I would like to see how well it would hold up to cooling as you increase power. Even if you could just give me the voltage and architecture your running in this build.
Lastly how much current are you putting through your peltier to cool the gas enough to condense it? I would be interesting to see if this is a bit of a power saver =)
@kyle2194 , It actually makes it worse to have a heat sink. With the boiling liquid, you are immediately sapping the heat directly from the chip as it boils. If you put a heat sink on it, then you lose this ability, and the chip has to heat *through* the heat sink first. (actually has an insulating effect) It will still cool, just not quite as effectively.
Weak analogy: If you jump into freezing water with an aluminum suit on Vs. bare skin, which would chill your body down most effectively?
How hard is that liquid to clean off if you have to do maintenance on the motherboard?
That is actually really cool :)
I'm wondering a few things:
How hot were the CPU VRMs?
Would mounting a large, flat piece of copper help?
Will you be repeating this test with higher TDP processors?(115w+)
wouldn't it work better if you put a heatsink on the CPU? once it got to 34 degrees C the bubbling would still circulate the liquid.
@malvcalv , Heh, heh, no we didn't drink it. I can think of a lot of things that are not toxic, that I still wouldn't want to drink. :-)
I've put my hands in it, smelled it, etc. (virtually no smell), and it evaporates off your hands almost immediately. A version of this almost exact same chemical is used in fire suppression systems for server rooms in place of Halon gas. There are some interesting videos of it putting out fires.
Imagine the temps if you put a heatsink with no fan on the processor :O
@CorvalentCorp That is interesting, thank you!
This is very coool. I wish i had that, but where is the GPU in this system?
I do not have a company but i would love to play games with that kind of cooling
Has this been tested with high end video cards? I am just curious how cool it would keep a 580 or a 6970.
@CorvalentCorp
That would be amazing! And would also help server operations stay within newer environmental guidelines. Has google called? :P