ECAM Could Revolutionize Liquid Cooling

At Hot Chips 2023 a startup called Fabric8Labs discussed a new technology the company is calling ECAM, which could improve cold plates in liquid cooling (AIO) solutions. In this video Adam chats with Dr. Ian Cutress from‪@TechTechPotato‬ about the details and when we could see this come to market.
Read more at fabric8labs.com/
Buy PCWorld merch: crowdmade.com/collections/pcw...
Follow PCWorld for all things PC!
-----------------------------­---
SUBSCRIBE: kzbin.info?sub_c...
TWITCH: / pcworldus
TWITTER: / pcworld
WEBSITE: www.pcworld.com
#cooling #liquidcooling #interview
00:00 - Intro
00:23 - How Cold Plates Work
02:39 - What Is ECAM?
04:37 - Gyroid Structures
06:51 - ECAM Is Cheaper
07:50 - Optimizing Heatsinks
09:10 - Can This Be Used For Other Purposes?
10:04 - Air Cooling
10:44 - When Is This Coming Out?

[@owlmostdead9492]

3:51 wrong laminar flow is exactly what you don’t want for heat transfer, you want turbulent flow. Laminar flow acts like some form of insulation and slows down heat transfer.

[@kuoster]

There are metal printed gyroid structured heat exchangers for helicopters and machineries, interesting to see this being used for PC cooling 😃

[@casfren]

4:10 I'm no expert, but I've heard that you don't actually want all laminar flow, because you run the risk of having only the surface water getting warm, but the centre of the water jet just pass by so linearly that it won't allow the proper transfer of heat. so having some turbulent flow that can impact against hot surfaces is more beneficial then only having linear flow. Once again, i could be wrong. fluid dynamics are hard.

[@absolutesadlad2297]

1:40 skiving is actually a cost saving measure and worse all around compared to CNC milling the fins. skiving, much like rolling actually decreases the size of the metal's granular structure to increase it's tensile strength. by doing so also decreases the electrical and thermal conductivity of said metal.

[@shanent5793]

It's about 3% worse conductivity and if they need even that back they can anneal

[@absolutesadlad2297]

@@shanent5793 depends on the metal. With copper it can be as high as 10%+ You also have to consider waterblock manufacturing. these coldplates arent being annealed after they are manufactured.

[@shanent5793]

If they're not annealing it's because it's not worth it. I don't see how skiving can be "worse all around" when it's clearly economical. Skiving doesn't produce chips, and hundred micron end mills would have to turn in the millions of RPM to be as fast. At conventional speeds you would be lucky to cut a single channel in a minute. If you really want to carve them use a metal shaper or a lathe

[@absolutesadlad2297]

@@shanent5793 thats why 95% of the market uses skiving to begin with is its economic impact. And why you only see a handful of companies actually mill their fins. Your average consumer doesnt care about a negative 3-6c reduction vs paying another $90 ontop of an already $80-$100 waterblock. But you also got what you pay for when you compare something like a XC7 versus a Heatkiller IV.

[@bug-hh8lk]

Awesome video -- I love this type of preview into tech for possible future products.

[@netoeli]

Noted I'll make sure to open it up and have a good look inside, thanks 👍

[@hukama6911]

Turbulent flow actually has better heat transfer coefficient

[@michaelglass1386]

My question... 1. How much of an increase in pressure drop is seen in a gyroid cooler ? With an increase in pressure drop a reduced flow is seen in the same pump. To offset the increase in pressure drop of the gyroid cooler you must increase the pump rpm to get the same flow as finned coolers. 2. With the increased rpm of the pump, how much additional noise is created by the increased rpm pump?

[@shadow7037932]

Huh interesting. I made something similar to this via PCBWay 3D metal 3D printing. Printed out of Aluminum though due to cost of copper printing. Used it to cool a high powered custom 98+ CRI LED setup for my workbench and it's basically silent with some Acrtic P12s running at very low speed and the LED temps stay around 40C.

[@FusionDeveloper]

This video addressed my concerns with turbulence, restricting flow and fragility.

[@lordofnothing.]

wow. thats very cool. lets hope Artic picks this up soon:) would be interesting to see how much of a temps reduction it would tranlsate to though.

[@spuchoa]

This was very informative, I hope this technology make it to market soon.

[@themusesquad8554]

How can You have the Doctor without Gordon? They are practically dating!! :)

[@lennylin-uf2yv]

It’s interesting that you have foreseen it happen 8 months earlier! We see Fabric8Labs launch the cold plate with Asetek in Taipei at Computex 2024!

[@jaymacpherson8167]

The strength of the Gyroid only needs to be enough to resist the pressure of the water flow, which is likely very low. The surface area can be far higher than linear fins, which is better for heat transfer beyond the mass of the copper. Additionally, it looks like flow would be turbulent, which also can increase heat transfer rate.

[@sabbilon4542]

Seems like a limiting factor will be the thermal conduction from the IHS through their metal structure. Thermal conduction is dependent of cross sectional area not surface area.

[@Longgshot]

What you want is turbulent flow, not laminar flow. Also, the cost comparisons the company makes are against 3D printing processes that are not commonly used, mainly because of that, cost. So their process might be cheaper to make the model they are trying to make, but i highly doubt that it is even remotely cheaper compared to what is commonly used today to make regular water cooler fins, which is skiving a solid piece of copper (or aluminum). And, if i'm undestanding this correctly, from the info that are on the graphs, by their testing they decrease thermal resistance by about 35%, but also reduce surface area by around 25% for the Gyroid 50%, and for the 80% Gyroid the reduction in surface area is around 70%. That doesn't seem to be a good outlook, the 80% one seems like a really bad idea no matter how you look at it, and the 50% one would be ok if the process was cheap, but by the looks of it, it is highly unlikely, since they are using a form of 3D printing, and 3D printing isn't exactly cheap. Even tho they are using a relatively cheap 3D printed method. It also looks like it would clog easily. So yea... i'm not exactly hopeful for this whole thing.

[@toddincabo]

👍 Is there a doctor in the house? Why yes.

[@digitalplayland]

A microfilter is going to be useful on the LC loop.

[@18earendil]

A microfilter is probably a good thing to add even to a classical water cooling loop.

[@leandrodrace]

That`s amazing! can`t wait to overclock some chips with a block like that.

[@Dogzilla07]

I wonder if these Gyroid structures could improve the wicking structure inside heatpipes for air cooling

[@Veptis]

My AIO radiator just died last month. I am building a new system with custom loop cooling in October. Interesting to see there is still innovation for water blocks. Roman showed that Alphacool and their newest generator of CPU blocks wins the performance comparison. And instead of sheers with channels, they also had pins in their design. Knowing that Intel is working on "3rd generation cryo Cooling Technology" which might drop with 14th gen Desktop (MTL being called Ultra seems to be confirmed due to GNA notice in openVINO 2023.2) - it would be interesting to see them go deeper than just water block with stacked Peltiers. Perhaps something like direct die as well. Intel also recently hired a specialist in computational fluid dynamics for their graphics research team. Could just be about HPC software. But maybe we hope they bring some interesting cooling tech to their workstation GPUs (to compete with RTX 6000 Ada) with build in blocks.

[@nexusyang4832]

Tech Tech Potato Chips!!! :D

[@C.J.G.]

6:51 Doesn't this just mean this ADM process is cheaper than other ADM processes? This doesn't mean its cheaper than drilling fins into a block of copper...

[@mgkleym]

Interestingly microfins are not drilled, they are made with a process called skiving where a blade shaves off a thin sheet of copper from the plate that is then straightened up into a fin.

[@TechTechPotato]

Yeah compared to skiving it's not cheaper, I kind of mispoke/misunderstood the question a bit, but skiving has a bunch of other issues to help drive down cost.

[@ledoynier3694]

yep. sintering and additive manufacturing is VERY expensive compared to conventional machining, and a lot slower. It's totally impractical for large scale manufacturing. It's usually reserved to small volume and more frequently for prototyping. That coldplate design must also be a ton of fun to clean once it's gunked up :) forget the toothbrush and get the pressure washer out, if it even survives the cleaning :p

[@IrocZIV]

I though to get heat to transfer from a material into a fluid, you want some turbulent flow, as a skin formed on the surface or something.

[@aaronhopkins7354]

Had the same thought, but it's been 20 years since I took Transport. :|

[@zuckdaddy1596]

depends on the nature of the cooler. I assume you're refering to early water blocks where the water was sort of forced through minature holes to generate turbulence, right? to my knowledge, current design philosophy is that you have micro-fins somewhat comparable to sandpaper which will sort of "catch" the water thereby forcing the water at a much increased flow rate while preventing that "skin" you mentioned forming. this is partially why AIOs (and blocks in general) tend to get worse over time. some of it is certainly simply aging pumps and lower fluid volume, but part of it is that those micro-fins are eroded away over a long enough period of time. usually years with decent stuff these days

[@georgejones5019]

Turbulence is also bad because it means you're not not having new liquid or air cycling through into system. Which can result in hot spots where it's not flowing out. It's also highly dependent on the design.

[@P4NCH1]

Yes, you will want a flow that tends to laminar or turbulent depending on your application and the characteristics of your structure. The one "mistake" I see on the video. And as a comment: I think it'd be interesting to see the tolerance of build up debris on that "swiss cheese structure". It will certainly be a lot less tolerant to tinted fluid than a fins one, for e.g.

[@TechTechPotato]

@@P4NCH1 In one of the slides it talks about the roughness - one of the benefits of the gyroid and the manufacturing method is that the roughness is on the order of 50 microns, rather than 200+ microns in fin structures. It's obviously a balance that's also mediated by the density of fins or gyroid holes.

[@kennethsanford9685]

I bet @derbauer would like to look into this

[@Phil-D83]

Wait for deepcool, etc to make a version

[@marcasswellbmd6922]

I can totally se those little holes getting cloged up quick..

[@justfasial01]

I find it hard to believe it's cheaper to 3d print those blocks vs the regular machined finned blocks.

[@seitenryu6844]

It's cheaper vs other additive methods, likely more expensive than subtractive methods(machining, milling, or skiving).

[@greebj]

Reliability and longevity too, surface contamination is a function of surface area, as is deterioration of the structures from even mildly corrosive flushes for maintenance

[@shanent5793]

So what, nobody's saying that they're cheaper, only that they're better optimized

[@justfasial01]

@@shanent5793 It's literally in one of the slides and the host said "Cheaper than this?" and Ian said Yes lol

[@shanent5793]

​@@justfasial01you're hallucinating, the slide compared it to incumbent additive manufacturing tech, not the regular machined finned blocks. When Adam said "cheaper than traditional methods?" the guest clarified "between different additive manufacturing technologies, it's the cheapest around"

[@mikhailstepanchenko7284]

Seeing the name of the company, does Der8auer have any connection to this? Also very wonder to see the real results

[@EyesOfByes]

Gyroid looks a bit like 2019 Mac Pro grill

[@Recycled]

That's super cool! 3d print the world🎉

[@WarisAmirMohammad]

Arent gyroids the gribbly dancy things from animal crossing

[@MrBisous34090]

no one uses this type of design because once clogged it will be impossible to clean it completely, so if one day you have a problem with oxidation reduction it's over, your block is good for the trash or to become a paperweight and if this happens during the warranty period it will generate a lot of return and cost the manufacturer very dearly

[@tacticalcenter8658]

Server manufacturer's don't screw that up like the brands your use to using.

[@DoNotFitInACivic]

A properly done closed loop hydronic system, boilers or chillers do not get gunk. Of course that requires a good number more steps than most enthusiast level folks know to do. We service decades old chiller systems that are still running their original coolant.

[@drew004jc]

seems like it would add more limitations to your build than it would be a benefit..............not really something for the guy who wants hot pink fluid on his hello kitty build............might work as part of an AIO design with a special fluid...........but why? .02 degrees or some inconsequential number

[@TheGreatAtario]

The future of liquid cooling is realizing that it is, and always has been, a meme

[@owlmostdead9492]

If you want performance along with silence and no throttling, it's still the only way to go. Silent = I can't tell if the PC is on from an arms length away, even under full load.

[@bobby0081]

Where is Gordon? Is he sick, or was he terminated? WTH?

[@TechTechPotato]

He's on medical leave.

[@bobby0081]

@@TechTechPotato thank you. I hope he's okay and can return soon.

[@shaytal100]

Good chance that I am dumb here, but if not the people that designed this are idiots: The heat capacity of water is higher than copper. The heat transfer of flowing water is also better than that of non flowing solid copper. There also should not be a real heat flow resistance on the junction between the copper and the water surface because they are touching each other. So there should not be the need to have a big surface area! Best cooling should be to use as less copper as possible and bring the water flow as close as possible to the surface of the chip. So am I dumb or is Dr. Ian Cutress and ECAM dumb? Honest question, I really think there is a good chance it is me :)

[@shanent5793]

There's a thin layer of water that isn't moving where the water meets the copper, so the cooling capacity is limited by the heat that can be transferred through that boundary layer. Since Cu is a better conductor than water, it's more effective to make that boundary layer as large as possible. A larger area also reduces the pumping losses. A single, thin channel the size of the die would have a lot of resistance to flow, and thus require more power than a wide channel. The tortuous path of the gyroid encourages mixing between the boundary layer and the water flowing down the centre of the channel.

[@shaytal100]

@@shanent5793 Just after I finished writing my comment I thought about exactly what you wrote. The water flow is zero on the Cu surface but grows (somewhat) linear with the distance to the surface. So the heat flow resistance is bigger than in copper for a thin layer, until the flow is fast enough to compensate for the lower heat conductivity of water. So to increase the overall heat flow one could increase the flow of the water and therefor make this thin layer of worse conductivity thinner or one could increase the surface. And increasing the surface should be favorable because the pump does not have to work harder (not much at least). Sorry I am basically repeating what you already wrote. Good points, thanks! And thanks for letting me know I am the dumb one here :) But in my defense it is a complicated matter.

[@Spark010]

Very interesting. If it works for fluids, it will work for air cooling too 😁

[@michaelglass1386]

Only if the pressure drop is small enough that the fan can push enough air through the cooler.

[@claucmgpcstuf5103]

Wel. Tha is a cool experiment yeaaas . Foca cool but it is probably to dance for all of this normal produredr.. etc Ian man find a we pu produset your self in lemitid nr and sel ti yous self before you panaten. If you can

[@Cxs1a3]

4raud

[@nanimonoda6897]

but it won't

[@p4radigm989]

liquid cooling for CPUs is and has always been stewpid water and PCBs don't go well together, it's always going to leak, always going to be a pain in the b-side

[@greebj]

I liquid cooled my laptop, custom built a heatsink using copper tubing, a cheap AliExpress copper block, and solder It runs a 9900k at 4.9ghz / 170W on a laptop on a 95W CPU heatsink For 5 years now One minor leak (from the detachable connectors outside the case, not from my diy fabbing), zero dead hardware Properly engineered & manufactured watercooling is easy you just have to know what you're doing I haven't had any cooling issues in my E34 and E39 BMWs either, it's not rocket surgery, it's just maintenance

[@Cutieplus]

AIO is pretty cheap and reliable today. Even those $30 dollar 240 AIO from China can last for 10 years. I have one that has been used for 7 years 7x24 and still working perfectly fine today. I have just replaced a new one this year just because I need a 360 AIO for a more power hungry 13700K.

[@ImDembe]

What if i told you no?

[@p4radigm989]

@@greebj if you water cool a Laptop it's not really a Laptop anymore since you created a Frankenstein monster that it no longer easy to carry around. what a silly idea.

[@p4radigm989]

@@Cutieplus Water Cooling is still stewpid. It's a pain in the b-side, especially when you want to move you PC. It's just annoying, lots of work, and gives not much benefit really except for people who are crazy and want to top 3d-mark scores online.

[@Deltium5683]

I really do NOT like Adam. Since when did a video support guy get qualified to interview proper technical people. Adam simply does not have any proper technical knowledge or training and he acts like a child. Can we go back to Gordon instead?

[@realkarfixer8208]

Gordons been having health issues and has been MIA for a while now. Lighten up, Francis.

[@justfasial01]

I wish he'd just let him explain instead of the constant interruptions lol

[@entropywins_]

Adam has never claimed to be an tech expert, in fact he's typically self-deprecating about it. We all miss Gordon but don't take it out on Adam.

[@TechTechPotato]

Adam did a great job here - I'm liable to waffle, and he kept me on topic and helped keep me concise. Otherwise you might end up with a 40 minute video saying exactly the same stuff. That's what a good host does, and he's doing an amazing job while Gordon is on medical leave right now.