My name is Jason Wallace, I made the video on youtube, and coordinated the proof of concept project. (Google Analytics tipped me off to the thread from the resulting site traffic.) Our company is all about keeping things ultra-reliable and in production for years at a time, and acting as an OEM manufacturer for industrial computer systems (medical, military, oil/gas, industrial automation, etc.) So while the stuff we do as a company, isn't really that relevant to what you guys are doing here (sheer overclocking speedy goodness)... this stuff is always fun to play with and talk about. And I'm happy to answer questions.
Disclaimer: I'm not an expert. I've spent time talking to some experts at 3M, wrote an article about it this, and built the system. So I've got some experience with it, but keep in mind, anything I say may just flat out be wrong. (Grains of salt and all that.) All that said, I'm happy to answer questions about this method of cooling and will address first, some of what I've seen asked on YouTube and in the thread here.
1) Efficiency. This was a very ugly, tape covered, simple proof of concept. For this experiment, it was not about efficiency. The TE cooler uses a lot of juice, and for a single system it's cool looking, but just not cost effective for day to day average consumer use. We're looking at this from the higher end perspective of having desktop computing power in sealed industrial systems that would normally only permit mobile computing (lower power chips). Such as deep under water (pressure), very dusty, very humid, very hot, very whatever extreme environment that would normally kill a computer system...
But this stuff get's really efficient when you consider the following senario: 3M did a test with only 200cc of Novec fluid that boils at 49C, cooling an array of heaters generating 4000watts of heat... which is roughly the equivalent of (30) i7940 CPUs. They used a liquid radiator system to condense the vapor, and successfully sustained the cooling level.
So what does that mean? Think about a container like a deep freezer, filled with blade servers, side by side, super tightly packed together in a bath of this liquid. They bubble while doing their various server things, making heat/vapor, which is condensed by a overhead radiator. The radiator uses standard cooling fluid to carry the heat to the outside facility air, where another radiator / fan removes the heat. No compressors, no $$$ HVAC systems, no fans, no noise, no mess, and no fires (this stuff is a fire suppressant, also used in place of Halon gas).
3M estimated that a full server rack, using 80W of fans per kW (1000W) of server power, ends up costing about ~$2800 per year in energy costs,*just to run the fans*. and that's only moving the heat from inside the case, to outside the case, so large expensive to run HVAC units then have to remove the heat to the outside air. Using the above described Novec cooling method, that cost is reduced to $123 per year, and takes the heat completely out of the facility, also elminating the need for the expensive oversized HVAC systems. It also eliminates all the fan noise, as well as the dirt and dust from the fans.
2) Why not put the TE cooler in the liquid and chill the whole thing?
We accidently tried that, and lost a lot of fluid during the test. Remember this isn't mineral oil. With this, bubbles = cooling. You want it to bubble. Effectively, this thing is spending energy cooling (by bubbling) only where cooling is needed. If you dip the radiator / TE cooler / whatever in the liquid, you are then suddenly trying to cool *EVERYTHING* the liquid touches (including the ambient air touching the housing) vs. just the hot spots that are producing heat levels above the liquid's boiling point. So the hot spots will still bubble regardless, but now, since your cooling power is being sucked up by the warm liquid, you are not effectively condensing that vapor. Pressure rises, temperature rises, bad things happen. (leak / rupture in tank, fluid loss, eventual temp control loss) By concentrating only on the vapor, you are now only removing the heat from the areas that need it.
3) Why not use a heat sink, and increase the surface area of the thing you are cooling?
To use a really bad analogy: If you jumped into freezing water wearing an aluminum suit, or with only bare skin, which would cool you off faster? Again, this isn't air cooling or mineral oil cooling... This stuff is boiling, and the boiling action is what carries the heat away. It's just boiling at 34C (93.2F). So as fast as the liquid can get to the chip, is as fast as it's carrying the heat away. An aluminum block, adds more material for the heat to have to pass through, and actually creates an insulating effect.
4) Why doesnâ€™t the fluid heat up?
It does. And as it passes the fluid's boiling point, it turns to vapor (evaporates). But since the entire board is not over the liquidâ€™s boiling point, you only really see bubbles where the hot spots are. The other general warmth just translates into a more rapid evaporation. (just like water evaporates more quickly when warm / hot vs. cold)
5) Would it work for graphics cards, etc.
Sure. The only problem I see with graphics cards, is that it wonâ€™t look as interesting. You would need a riser card to allow the card to lay down over the motherboard, so you need less fluid volume (unless your budget allows for a lot more fluid) vs. a vertical card standing up off the board. This would cover up many of the interesting areas on the board, making the whole thing less entertaining. From a purely functional standpoint, sure, it would work fine. An fun exception to this would be if you made a custom acrylic case, with an area that sticks out the front, just big enough to contain the graphics card. This would double as a stand to keep the tank upright/stable, and allow the card to work without a riser card or similar modification.
6) Will it work with hard drives?
Nooooooo. Hard drives are mechanical, and depend on a cushion of air between the heads and the platter, as well as precise timing and movements of the internal parts. Putting them in the liquid would cause very bad things to happen to your drive. Hard drives have vents to allow air pressure to equalizeâ€¦ I suppose if you sealed those, maybeâ€¦ but youâ€™re asking for trouble. Solid state hard drives however would be perfectly fine.
7) Long term effects:
No known long term negative effects are known for the boards, caps, or components. Exception: Plastics with a lot of plasticizers (stuff that makes plastics very flexible). Over time, the chemical will leech the plasticizers out of the plastics. Iâ€™m GUESSING that this would result in more brittle, firm, plastics. Ideally once you have everything routed, over a long time, the cables would sort of â€œfreezeâ€ in place, and you might have to be careful not to flex them too much when opening the case to work on the PC. I donâ€™t think it would have a significant impact on the cooling capability of the liquid.
8) OMG the stuff is expensive
Yeah, it is. $300 a gallon, plus or minus 20-ish bucks. Itâ€™s a very precise, specialized industrial chemical, intended for industrial use. Sorry. Seal your housings well. A leak will literally make money evaporate. The good news is, if your housing is sealed well, you should almost never need to mess with it again, no need to add fluid, etc.
9) What if water gets in there?
Iâ€™m told this particular flavor of the liquid absorbs water. Some water wonâ€™t really mess with stuff. If you dumped a gallon in with a gallon of the Novec, yeah, that would be not goodâ€¦ and it would probably suck for the PC. I donâ€™t know the ratios of how much is too much. Since the housing is sealed to prevent fluid loss, you arenâ€™t going to pick up much water from the atmosphere. Other versions of the fluid will reject water, and it remains separate from the fluid.
10) What if itâ€™s used in a very hot environment?
Then you would use a version of the fluid with a higher boiling point, or the liquid will begin to pressurize the container. Figure that your CPU temp will always stay around 15-20C over the boiling point of the liquid... and go from there.