3M Demonstrates PC With Submersion Cooling

AlphaAtlas

[H]ard|Gawd
Staff member
Joined
Mar 3, 2018
Messages
1,713
A 3M employee demoed a CPU cooled with a low-boiling-point liquid years ago, and since then, some server manufacturers have shown off similar systems. But this year, that same 3M employee decided to build a more compact version of the original demo.

Check out the final build here.

At least some of the previous demos used 3M Novec 7200 or 7100 fluid, which have boiling points well below 100C. When under a stressful workload, the CPU's IHS boils the volatile liquid, which is re-condensed and cooled by what appears to be a 360mm radiator. Like the previous demo, the entire system is sealed, but it seems that 3M hopes manufacturers will make such cooling systems more accessible in the future. Fuzzie documented the entire build in a series of videos, and it looks like this isn't the end of his experimentation.
 
I remember first seeing submersion cooling in 2004 and being fascinated by it. I think it's too impractical for home use, but cooling performance is impressive.
 
So it's not enough that the system is watertight, it has to be airtight as well? Nah, no thanks.
 
Interesting that the motherboard is just dumped in the Novec stuff.
Then there's a radiator inside the tank being cooled by a radiator outside the tank.
Seems like that part is kinda inefficient?
I guess I just want to see that Novec stuff in some sort of sealed system.

Also would have been cool to see how it works without the heat spreader. Do you need one in this circumstance?

Neat stuff.
 
Interesting that the motherboard is just dumped in the Novec stuff.
Then there's a radiator inside the tank being cooled by a radiator outside the tank.
Seems like that part is kinda inefficient?
I guess I just want to see that Novec stuff in some sort of sealed system.

Also would have been cool to see how it works without the heat spreader. Would love to see different parts of the die vaporizing the Novec differently.

Neat stuff.
 
I wonder if performance is improved or degraded if a small heatsink were added to the CPU, thereby providing more surface area for the fluid to boil off of.
 
I wonder if performance is improved or degraded if a small heatsink were added to the CPU, thereby providing more surface area for the fluid to boil off of.

IIRC someone tried this using a naked die as well.
 
I wonder if performance is improved or degraded if a small heatsink were added to the CPU, thereby providing more surface area for the fluid to boil off of.
As counter intuitive as it sounds, adding a heatsink may actually make it perform worse. You actually want the liquid to boil because the phase change from liquid to vapor takes a lot of energy (so-called latent heat) which comes from the heat source (in this case, the CPU). This enhances the cooling significantly over normal conductive heat flows. The heatsink would require the CPU to get a lot hotter before the boiling point of the liquid is reached, so might be less effective. I'm sure there an optimal surface area for a specific amount of heat, so some sort of spreader might be helpful, but it would depend on the CPU/chip you are trying to cool.
 
^ exactly the same as why heat pipes are ubiquitous in high-performance CPU/GPU cooling, the heat transfer from phase change is really really good.

Also, remember the application 3M is demoing is for servers, where rack density and external costs may be greater than an incremental cost of essentially building giant heat pipes around the entire server hardware.
 
I recall using mirenal oil on an old AthlonXP 2500+ and a geforce 6800GT like ages ago.

Crude experiment I might say as I left the fans. They would spin slowly. On the plus side it smelled like french fries :D:D
 
how low boiling point ? i have sailed with Olympic pool amounts of chemicals on tankers, and most of them had a boiling point well under 20 degrees as the remains in the connector was "boiling" when we disconnected and this was in early spring with barely over freezing temperatures, also most of them would give you cancer just by uttering their name.
 
I immediately though Anomalous Materials lab.

Also, I'm still wondering how 3M went from making floppy disks when I was 10 to making high visibility jackets and stickers and adhesives now.
 
I immediately though Anomalous Materials lab.

Also, I'm still wondering how 3M went from making floppy disks when I was 10 to making high visibility jackets and stickers and adhesives now.

The company is literally called Minnesota Mining and Manufacturing.

Their history is incredibly diverse as you can tell just by the name.
 
Wow........
Nothing new. I recall seeing similar demos of Novec fluid 10 years ago. Back then it was several hundred bux a gallon. I see nothing has changed.
 
That's about how much you end up spending just in damn fitting for a custom water loop...

I would expect that you would need more than 1 gallon for this in most cases.

Edit: Although this small version from the video may get away with a little over 1 gal.
 
I would expect that you would need more than 1 gallon for this in most cases.

Edit: Although this small version from the video may get away with a little over 1 gal.

lol fine, get a second gallon for the cost of your blocks and tubing.

(looks like radiators, fans and pumps are still necessary for this stuff, so it's "only" of similar cost to a custom loop if you stop at 2 gallons)
 
This would be super easy to make. Get a nice copper box to hold the rig put rig inside braze a top on it with a fill port, put a vacuum fitting on it and a emergency pressure relief. Pull you a light vacuum to make sure it's air tight and fill the box up and let it rip. If you ues enueno copper it should act like the heatsink.
 
This looks like it would be a fun DIY projects.

Last time I looked up that Novec stuff it was pretty expensive though :(
 
Interesting that the motherboard is just dumped in the Novec stuff.
Then there's a radiator inside the tank being cooled by a radiator outside the tank.
Seems like that part is kinda inefficient?
I guess I just want to see that Novec stuff in some sort of sealed system.

Also would have been cool to see how it works without the heat spreader. Do you need one in this circumstance?

Neat stuff.


Why wouldn't one just pump the fluid directly into a radiator rather than doing that dual radiator nonsense?

Maybe they are trying to make it boil to absorb the heat?
 
Yes it boils at 61C.

Hmm.

At most fan settings my custom water loop keeps my GPU in the 30C range. In the mid to high 20's if it is cool in my office, and I have the fans maxed.

Boiling at 61C, even though probably very effective at removing large quantities of heat at that temperature, seems like it would result in higher temps than a good water loop.
 
I expect it will begin evaporation significantly before the boiling point.

I guess so.

Otherwise it doesn't make sense.

I was just able to watch the video, and it's bubbling like mad around the CPU socket, but the temp listed is in the high 20's, not 61...
 
PV=nRT still applies, guys. Slight vacuum and the boiling temp drops a decent bit. All depends on where you want the sweet spot of your evaporation/condensation cycle to be at.


Good point. I assumed it was operating at atmospheric pressures.

Accurately controlling pressure seems a little much for home PC cooling, but maybe for data centers?
 
I'm speculating as much as you guys! I can see a case for some sort of wholesale phase change heat exchange system, given the HVAC and power density scales that server farms need, even if that is utterly ridiculous for the home user. I just don't have the numbers or the engineering (not exactly my specific domain of expertise either).
 
Back
Top