3D-printed fan-less and pump-less liquid cooler can deliver 600 watts of cooling for data centers — passive design provides reusable heat, exceeds project performance expectations by 50%
Better watch out, Frore.
Get 3DTested's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Move over, Frore, there's a new player in town. A research team comprising boffins from the Danish Technological Institute and the Heatflow company have come up with a 3D-printed cooler that can draw a whopping 600 W off a chip fully passively, without any pumps or fans, 3D Printing Industry reports. The design is naturally aimed at datacenters first and foremost, though one can imagine a world where variations on the design could work on desktop PCs and workstations.
Should the presented figures pan out, to say this cooler is impressive is quite the understatement. The original specification set an already-ambitious target of 400 W, a figure that was exceeded by a nice, round 50%.
Other than a presumably exquisite internal layout, this heatsink doesn't seem to have any special tricks. It's a standard two-phase thermosiphon, where coolant heats up and evaporates at the bottom, and condenses into vapor at the top, coming back down via gravity alone. The part shown in the picture is the heatsink that will connect to a radiator somewhere without any moving parts needed for the heat transfer.
The cooler's dissipation ability is impressive enough, but the liquid that goes out of it is claimed to be at 60 to 80°C, making it easy to recover and use in other heating networks for a double-whammy. The report claims these figures are superior to standard datacenter cooling that whisks away heat at lower temperatures, making it harder to reuse.
Seeing as the cooler is 3D printed, it has very little material wastage for production and only employs one material, making for efficient production and easy recycling. The underlying chip temperatures should also be lower than with conventional designs, helping with their longevity — a boon for AI accelerators that tend to live fast and die young, with exceedingly high failure rates.
The project's budget is arguably even more surprising than the end results. All it seemingly took was 10 million Danish Crowns (DKK), a figure that translates to $1.56 million, or 1.34 million euros. We do hope this technology makes its way to our gaming PCs... Whenever we can afford RAM again.
Follow 3DTested on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
Get 3DTested's best news and in-depth reviews, straight to your inbox.

-
Dementoss Reply
RAM is no problem, I've still got two kidneys...Admin said:We do hope this technology makes its way to our gaming PCs... Whenever we can afford RAM again. -
bit_user Nvidia's H100 (SXM) already surpassed 600W in 2022!Reply
If they intend to use it on CPUs, then it still has a potential window of relevance. -
thestryker 3D printed blocks for LN2 provide the best results possible, but also cost significantly more money. I don't see how the TCO on something like this balances out as that would be the only way it really makes sense. As bit_user points out at a rack scale this design would only be sufficient for CPUs which makes it even more questionable.Reply -
bit_user Reply
Although my post was rather glib, I don't know what were the project's objectives. $1.56 million seems like quite a lot of money for a simple research project, but perhaps this is just the first phase in what's ultimately meant to be something that can be produced at commercially-relevant costs & volumes.thestryker said:3D printed blocks for LN2 provide the best results possible, but also cost significantly more money. I don't see how the TCO on something like this balances out as that would be the only way it really makes sense. As bit_user points out at a rack scale this design would only be sufficient for CPUs which makes it even more questionable.
I definitely support more efficient & reliable cooling. It seems to me this should fall somewhere between a water cooling system and a heat pipe, in terms of reliability. -
alan.campbell99 This certainly is an interesting development and more efficiency is great, however, I must say I'm more jaded these days about high-end tech news. I used to follow, among other aspects, HPC-related news with genuine interest because it was bleeding edge stuff going into supercomputers and the like. Research and science-oriented developments. It wasn't anything I'd ever expect to touch but I thought the progress was cool.Reply
For now though it seems it's all about the AI datacentres, for me it leaves a bitter taste given the various related issues with these.
Perhaps it could be adapted for the desktop. -
thestryker Reply
Oh certainly, but the financials have to make sense and at the datacenter level I just don't see it. Liquid cooling cuts down enough on operational costs, despite being more expensive up front than air, to make it viable. If the block costs here are much higher and the only benefit is losing the pump then the operational cost isn't going to be dramatically different than liquid.bit_user said:I definitely support more efficient & reliable cooling. It seems to me this should fall somewhere between a water cooling system and a heat pipe, in terms of reliability. -
reghir Reply
I have to admit that looking at the 3d print I thought it looked like an amoeba mated a partial octopus:)Admin said:3D-printed passive cooler can pull 600 W off a datacenter chip
3D-printed fan-less and pump-less liquid cooler can deliver 600 watts of cooling for data centers — passive design provides reusable heat, exceeds...: Read more -
Charles Cabbage Thermosiphons are extremely cool; I'm eagerly awaiting the Noctua consumer offering, even if it costs an arm and a leg.Reply
Https://www.tomshardware.com/pc-components/liquid-cooling/noctuas-futuristic-thermosiphon-cooler-is-back-and-bigger-than-ever-at-computex-2025-but-still-no-closer-to-release