This server lives underwater!

[Guest]

Cooling is an expensive business for servers and datacenters, often involving HVAC. As CPUs and GPUs pull more power, more exotic solutions are needed – Enter immersion cooling!
 

 

[HowardPlayzOA]

That's why the forum site was down.

[dogwitch]

5 minutes ago, HowardPlayzOA said:

That's why the forum site was down.

nah... it was coltan getting revenge. on linus firing him again!!!!

[costafilh0]

@GabenJr you forgot the link to the MSI stuff and giveaway 

[OroBoros86]

I wonder if this can be bought commercially, looks cool though to build an entirely enclosed sffpc with a radiator for it.

[Joshua_P_Gervacio]

I'd really like to see a video where you guys use this on off the shelf components to see if this is a viable cooling solution for the average consumer. Also I was wondering if the solution is available to the general public to buy.

[Nystemy]

To a degree I have to be very very doubtful of their claims. Especially the "safe" part.

 

The biggest red flag is that they don't provide the safety datasheet without registering on their website. _{(A lot of other companies like 3M does provide safety datasheets publicly, not providing it publicly is just a sign of it not being particularly good...)}

 

The supposed non flammable part indicates that it likely is a fluorinated compound, and so far practically all fluorinated hydrocarbons have been shown to degrade the immune system to a sufficient degree that these compounds are getting phased out of the industry. _{(just like chlorinated compounds have been for a couple of decades now.)}

 

 

Though, also have to question the thermal conductivity numbers in the video.

Since water has a thermal conductivity of about 0.609 W/mK, not 300 to 2100 W/mK.

Mineral oil meanwhile is at 0.1 W/mK.

Air is about 0.025 W/mK at 1 atmosphere. _{(To get up to 100x better thermal conductivity one has to pressurize it by quite a bit. Since yes, thermal conductivity in gases is proportional to pressure and why one gets bugger all thermal conductivity through a vacuum thermos.)}

 

So the numbers in the video are completely incorrect by many orders of magnitude.... _{(Also don't know where the "2" in W/mK came from. Since it isn't w/m²k)}

 

I should probably add that I don't at all expect the "SmartCoolant" to reach anywhere near 2050 W/mK in practice.

_{(Also for everyone now questioning these low numbers, then remember that liquid/air flow makes a huge impact on performance. Since the mass of coolant won't just sit there still and soak up the heat, it will move both due to convection, but more so from pumping. But this is also why laminar flow is abhorrent for cooling performance since then one is back at thermal conductivity being the main limit again since the skin of warmer fluid won't really conduct much into the rest of the fluid, having turbulence to mix up the flow is a major advantage, even for air cooling. Though, turbulence in air cooling tends to generate a hissing sound at higher airflows, something most people don't like.)}

[PainfulTruth]

The other details published in the video seem to line up with the tech specs of SmartCoolant. See images below.

 

 

I also retrieved the full technical specifications, for anyone that is interested (doesn't list much more than the abbreviated version) - see attached.

 

@Nystemy, I wonder if just manipulated the thermal conductivity to their own novel units. It wouldn't be difficult to determine the 'thermal conductivity' as a factor of area [m²], instead of length [m]. Maybe they were trying to hide the fact it is SmartCoolant.

 

I would be interested to know the price and whether it is available to the public. Although, synthetic chemicals are quite expensive.

Submer_SmartCoolant_Leatlet_low.pdf

[Nystemy]

3 hours ago, PainfulTruth said:

I wonder if just manipulated the thermal conductivity to their own novel units. It wouldn't be difficult to determine the 'thermal conductivity' as a factor of area [m²], instead of length [m]. Maybe they were trying to hide the fact it is SmartCoolant.

Thermal resistance is actually not measured in length, but rather volume.

 

But as seen bellow, this volumetric aspect disappears from sight _{(But it is still there!)} when we simplify the formula, since the formula is:
Watts / (Width _{[in meters]} x length _{[in meters]} / thickness _{[in meters]}) x delta Kelvin
Simplified by doing:
Watts / (meters x meters / meters) x delta Kelvin = Watts / meters x delta Kelvin
 

The formula has just been a bit oversimplified. This is very common with a lot of physical formulas and a lot of caution has to be considered when working with physics, else one will stumble into beginner mistakes. This is especially common around volume, since if an observed effect is inversely proportional to one axis but proportional to the other two, then we end up with a "single axis" even though we in actuality still need all three for the formula to work.

 

Further reading for those interested in how this whole watts per meter Kelvin thing actually works and what it means then lets continue.

 

The formula simply describes how many watts we need to have flowing through the interface to have 1 Kelvin (or C) temperature difference between each side of the interface.

 

If we have a block of copper with a thermal resistance of 401 W/mK.

And this block is 2 x 2 cm in surface area on each side and 1 cm thick.

Then how many watts do we need to have a 1 Kelvin temperature difference over it?

The answer is: 401 x 0.02 x 0.02 / 0.01 = 16 watts

 

If we shave the block down to 1mm thick.

Then we need: 401 x 0.02 x 0.02 / 0.001 = 160 watts

 

If we have a thermal compound instead with a resistance of 8 W/mK

And our interface is 4 cm x 5 cm and a thickness of 0.1 mm.
Then we need 160 watts to get a 1 degree C difference over our thermal compound.

[ElReyLionel]

On 8/17/2022 at 5:59 AM, Joshua_P_Gervacio said:

I'd really like to see a video where you guys use this on off the shelf components to see if this is a viable cooling solution for the average consumer. Also I was wondering if the solution is available to the general public to buy.

A straight up 'Underwater' build variations video would be wicked. In typical Linus fashion he could probably split it up into multiple videos for that bread, but I think it would be very cool to see how cheap you could make something "Underwater" and not compromise too much on performance. With proper cleaning, I can image anything becoming a computer case at this point, with this possibly becoming the easiest way to make a tank and just drop a bunch of motherboards in with different builds. Would that even be possible? The questions honestly are endless since it is a new approach, perhaps dropping server racks into these tubs in a small HVAC room could be the key solution to temperature management, and no more cold control rooms, but is it even possible to begin with. I feel as if you are able to drop multiple computers in the tub, that would be a big leap in temperature management. 

 

Also, do you know how many fishtanks are literally out to rot? AND CHEAP?! I think LTT Labs should bring us the answers !!

[FoesPause]

12 hours ago, PainfulTruth said:

The other details published in the video seem to line up with the tech specs of SmartCoolant. See images below.

 

 

I also retrieved the full technical specifications, for anyone that is interested (doesn't list much more than the abbreviated version) - see attached.

 

@Nystemy, I wonder if just manipulated the thermal conductivity to their own novel units. It wouldn't be difficult to determine the 'thermal conductivity' as a factor of area [m²], instead of length [m]. Maybe they were trying to hide the fact it is SmartCoolant.

 

I would be interested to know the price and whether it is available to the public. Although, synthetic chemicals are quite expensive.

Submer_SmartCoolant_Leatlet_low.pdf 548.45 kB · 6 downloads

A heat transfer coefficient is not the same thing as thermal conductivity (or resistivity). Since they header'd it as such, I would assume they did actually mean the heat transfer coefficient (typical ranges for this kinda thing anyways). I imagine not a "beginner's mistake" on their end :).

 

This is used for things like line heaters or heat exchanges (which is pretty much exactly what this is) and the like. Think the U-value/overall heat transfer coefficient.

Thermal conductivity is involved in it, but so would be the convective heat transfer coefficients and things that affect liquid films/layers at the surfaces. And WAGs at fouling factors for what you'll expect to actually see. You can change U by fooling with flow rates, materials, geometry, etc.

 

W/[(m^2)K], Watts per metre squared Kelvin, would indeed be appropriate units.

[Psittac]

 

[Joshua_P_Gervacio]

13 hours ago, ElReyLionel said:

A straight up 'Underwater' build variations video would be wicked. In typical Linus fashion he could probably split it up into multiple videos for that bread, but I think it would be very cool to see how cheap you could make something "Underwater" and not compromise too much on performance. With proper cleaning, I can image anything becoming a computer case at this point, with this possibly becoming the easiest way to make a tank and just drop a bunch of motherboards in with different builds. Would that even be possible? The questions honestly are endless since it is a new approach, perhaps dropping server racks into these tubs in a small HVAC room could be the key solution to temperature management, and no more cold control rooms, but is it even possible to begin with. I feel as if you are able to drop multiple computers in the tub, that would be a big leap in temperature management. 

 

Also, do you know how many fishtanks are literally out to rot? AND CHEAP?! I think LTT Labs should bring us the answers !!

I guess the lab has their work cut out for them . Well assuming they can get access to the liquid that Anthony showed in the video. Though not to be a spoil sport, he did mention that the mixture was top secret so its highly unlikely that the company will make it available to the average user. But if they are confident that no one can reverse engineer their mixture, they might sell it to the general consumer.