Thermiderp

I've in fact seen the video before and its tentative result of possible small gains over water cooling picked my interest given that Linus has been interested in and made numerous videos out of digging out the best of the best performance of a given machine. See for example the newest video with Linus & Alex where they just liquid metal the Macbook pro; maybe this project should be though of as an analogous upgrade attempt over regular water cooling instead of a "new revolutionary technology". Secondly I feel that the rough lab setup in the referenced video is barely applicable due to the use of the water bath and the large bottle for both reservoir and heat exchange simultaneously; the glass is an insulator and secondly the liquid has quite little surface area through the bottle with the bath in comparison to its volume. I'd really like to see this built properly with a radiator, since neither the water nor liquid metal cooling performance is anything close to ideal in the video. And to address your statement of no academic value for a video from this idea, which might be somewhat true, I'd like to compare this project to the DIY air cooler / water cooler projects that Linus & Alex and Linus & Luke (in the Scrapyard Wars season N) have done, which have been really interesting due to the technical hurdles they've had to overcome. The principal reason to pitch this idea here at all is that I'd like to see these guy's serious attempt to do this well, even if the gain over regular water cooling is minimal. It'd be a somewhat like of an unicorn project like the mineral oil cooled machine that made a nice build log but which obviously didn't yield any academic gain. However, I'd argue that there could be value from their discoveries regarding the technical execution of the build. Though I'd still be optimistic that the liquid metal cooled system could be observably better which would be interesting to verify; of course it's not going to have groundbreaking results differing from regular water cooling due to the air cooled radiator but it could be somewhat more efficient and maybe the "academic value" would be in the attempt of pushing air cooling to its ultimatum, to see where the limit of air cooling is?

See my recent post where I discus Galinstan which has a dynamical viscosity of 2.4 mPa·s making it almost as runny as water. I feel that it should be quite pump-able?

Thank you for your replies. I do appreciate that the air cooling of the radiators might be heavily bottlenecking the system. I however feel that it would be interesting to verify this in action with substantially more thermally conductive coolant. The more thermally conductive coolant would be quicker to transfer its heat into the radiator possibly allowing it to reach lower temperature before leaving the radiator. Secondly I'd like to state that the idea of this cooling system came to me after watching the whole room water cooling vlogs, which I throughly loved watching for the amazing train wreck that it was. I feel that this build would make a really interesting video regardless of it being more effiecient in practice than a regular water cooled system. Edit: Additionally I'd argue that the more thermally conductive (and lower heat capacity) coolant would reach a higher temperature at the coolant blocks, making the radiators have a higher temperature where they'd be more effective relative to the ambient temperature. Though whether this would lead to any benefits in component temperatures in practice isn't clear.

What was the reasoning for the insufficient power of pumps? I did some research into the viscosities and usable temperatures of low melting alloys of Gallium and I ran into Galinstan which seems to match the physical properties of Conductonaut quite nicely. The given (dynamic) viscosity of Galinstan is 2.4 mPa·s whereas water's dynamic viscosity is 1.002 mPa·s, both at 20°C. Comparing these values for example to motor oil for which Wikipedia states "a typical motor oil could have a viscosity of about 250 mPa·s" which suggests to me that the desirable alloy of Gallium is comparatively just as viscous as water and so the viscosity alone shouldn't that much of an issue? Second part of the research tried to probe the cost of filling a loop with the liquid metal. Purchasing a commercial liquid metal would obviously be ridiculously expensive so I found Galinstan while trying to figure out an alloy that one could make by theirselves. Given the expected cost of the coolant I'd suggest that the loop be built as small as possible, maybe requiring a liter of the coolant at most in order to keep the cost in check. Supposing that we'd need roughly a liter of the coolant (assuming the typical Gallium Galinstan given in the wiki, 70% Ga, 22% In, 8% Sn here) we'd need roughly 4.5-5 kg of Gallium, around a kilo of Indium and ~0.5 kg of tin. I found a kilo of Gallium online for $299, and a kilo of Indium for $465 and assuming tin is cheap here, I estimate a total cost of the raw components of the coolant to be in the ballpark of $2000, which might be "cheap" enough to do an experimental video with. Obviously one will need to scale the $2000/1L of coolant if the loop needs to be larger quickly making the budget more hefty.

At least Conductonaut is liquid down to 10°C so it wouldn't need pre-heating at room temperature like pure gallium and so if would be much easier to use. If the metal would turn solid it would be pretty hard to get it working again. Mixing the liquid by theirselves is a good idea however, though it probably requires good know how to get similar properties to Conductonaut.

I'm hoping they can arrange something with Thermal Grizzly / another brand in between their $100k PC build projects. Surely way beyond the scope of any regular bloke though.

I'm writing this post to convince you dear reader that a liquid metal liquid cooled computer would be an exiting and suitably insane project for our cooling scientists at LTT to take on. The Why The thermal conductivity properties of liquid metal would be exiting to see in action in a liquid cooling loop. Imagine a heat pipe that extends all the way from the CPU/IHS to a radiator, but one that actually constantly circulates a cooler part onto the CPU. Mercury was my first thought but that is toxic and somewhat unsafe to handle so that's a no go (and not very attainable in larger quantities). But that isn't a problem at all since in fact the thermal compound alternative liquid metals are more or less alloys of Gallium, Tin and Indium and so they are safe to handle and readily available (in tiny quantities). The usable temperature range of these liquid metals is also excellent for this kind of an application. Only challenge is to get enough of the stuff to fill a loop with. The commercial liquid metals are mostly gallium so their specific heat capacities are around 0.370J/(g*K) (and at best since for example tin and indium have lower specific heat capacities) while the density of gallium is 5.904g/cm³ , so the relevant effective ability of the coolant to carry heat is 2.19 J/(K*cm^3), i.e. changing the temperature of one cubic centimeter of the liquid metal by 1 Kelvin takes 2.19 Joules. Water has the specific heat capacity of 4.2J/(g*K) and its density is 1g/cm^3 so the effective ability to carry heat per unit volume is 4.2J/(K*cm^3) and therefore the liquid metal's capacity to carry heat in the loop is lower by about a factor of two. On the other hand the thermal conductivity of liquid metals is outstandingly higher than water's: gallium has thermal conductivity of 29 W/(mK) though for example Thermal Grizzly Conductonaut is stated to have a thermal conductivity of 73 W/(mK) and water's is 0.598 W/(mK). So gallium and Conductonaut are 48.5 and 122 times as thermally conductive as water, respectively. The exciting feature of such a cooling loop is the outstanding thermal conductivity making heat exchange at the blocks and radiators highly efficient. Especially radiators will be superbly effective for shedding heat from the liquid metal coolant. Issues/Challenges: Corrosion. The gallium in the liquid metal reacts chemically with both aluminum and copper. The reaction with aluminum is strong enough to destroy any aluminum it comes into contact with quite quickly whereas the reaction with copper is less destructive as the gallium slowly seeps into copper making an alloy. All in all aluminum is a total no go and copper could be and probably is suitable at least for a somewhat short term experiment. Luckily it looks like copper radiators might be available so maybe the metal parts of the loop can be copper. Second option is nickel which is much more inert for gallium and that's why liquid metal is comparatively well suited paste replacement between the processor die and integrated heat sink since IHSs are plated with nickel. Nickel plated steel or other parts could be an alternative for copper. Aside from corrosion another challenge could be the weight of the liquid metal for pumping since the liquid is around 6 times heavier than water. Though I'm not at all an expert enough to tell if this is even going to be a problem. Secondly the viscosity and surface tension properties of the liquid metal will be different as well but might be that they won't be problematic either. It might be that a larger stronger pump will solve these simultaneously. Lastly the pump can't have aluminum parts exposed to the liquid metal either. Some thoughts on execution: Plastics should be alright so regular tubing strong enough to handle the weight of the liquid would probably be fine? Clear tubing so we'd see the metallic coolant in action. For maximum insanity do hard line tubing made of glass, though this might be too big of an unnecessary complication for this project. Diameter of the tubing and the viscosity of the liquid metal could be something to consider, maybe thicker fluid requires larger tubing? The coolant is so thermally conductive that it would probably conduct some heat upstream as well in the loop so it would be interesting to see some temperatures measured form different parts of the loop the see whether this is a big effect. Shouldn't hurt the performance of the cooling loop too much I don't think. Increasing flow rate would help mitigate this. So what do you guys think, is this idea worthy of a video / build log? Did I miss any critical issues that make the project much harder? I'd really love to see this made into a video, is this the best place to pitch the idea? TLDR: Linus et al. should build a purpose built liquid cooling loop and fill it with liquid metal (Conductonaut?) and see what happens.