We open the 7900XTX Vapor Chamber

[@eliotcougar]

It looks like sometimes during horizontal operation too much of the liquid gets trapped on the opposite surface... It would be interesting to try "inverted horizontal" orientation...

[@stevecade857]

My thoughts as well. We all assume gfx cards and their coolers are not dependant on the coolers orientation. Now seeing inside a vapour chamber it seems gravity could have a big part to play especially in the areas without any material offering capillary action. Surely gpu manufacturers have looked at this in the past and decided the best place to mount the coolers on cards. However, having seen their recent disasters I'm not convinced.

[@haakoflo]

@@stevecade857 Exactly. Btw, this is the same for CPU coolers, too. You don't want the cpu to face up if the cooler relies on heat pipes. Capillary action may only "push" a limited amount of liquid per second against gravity, and it is also limited by distance. Kind of how the flame of a candle is of relatively constant size whether the weather is colder or hotter.

[@lennyvalentin6485]

@@stevecade857 Regular heatpipes are definitely affected by gravity; for example PC chassis which have the card slots oriented vertically with the card slot brackets and video connectors turned upwards leave a GPU hanging vertically, kind of like a resting bat. This will cause heatpipe working fluid to amass in the lower end of the cooler, away from the GPU coldplate area where you want the fluid to be, thus increasing the risk of a heatpipe stall (IE, where the heatpipe runs dry in the hot end and basically stops transferring heat.)

[@MsTatakai]

AMD failed on testing because the benchtable is horizontal monted and computer cases is vertical mounted... maybe?

[@Stormkez]

@@MsTatakai was just thinking about that too

[@Duvoncho]

Ha! I knew you wouldn't be able to resist cutting into it. Fantastic stuff 👍

[@shadowarez1337]

Was waiting for this when this came up possible drying out I'm like yea he foingyto cut it.

[@stuartmorgan3654]

Twas never a matter of if, only when.

[@PSYCHOV3N0M]

EVERYONE could see this video coming. If you think otherwise, you're delusional.

[@Madness801]

"Didn't have time to do it yet" its not about resisting just time

[@davidepannone6021]

Thanks for everything you've done and still doing for all of us consumers, Roman. Happy new year buddy.

[@cryo_vL]

This is next level KZbin material. Like in the old days!!! Love it. Thank you for educational video.

[@phillipocanya6775]

in this case he did what in particular? and wash your brown-nose

[@neondemon5137]

@@phillipocanya6775 Calm down, buddy.

[@der8auer-en]

Thanks a lot :)

[@clarkecorvo2692]

@@phillipocanya6775 dude wanted to say something nice and you automatically feel the need to be a d*ck? yikes man.

[@M0nkNZL]

The video I have been waiting for! Thank you for being transparent and helping all of us figure out this puzzle.

[@der8auer-en]

thanks :) not much expertise for this specific topic but could be one more piece missing in the puzzle :D

[@kevinerbs2778]

@@der8auer-en I'm pretty sure the amount of water inside the vapor chamber the volume is too high. The reason for this is when you open milled into the vapor chamber the air pressure should have cause a lot of it to escape instantly. This is from the change in pressure as it's under less presure than normal atmosphere. The fact you were able to blow some of it off is proof. The water is pooling in cooler spots because the volume of water is too high, it then is away from the heat load as it condense elsewhere. You can see this when you blow off the water at the edges of the heat sink & just about 0 is near the gpu hot spot. When oriented side ways all the water drops to one side when it should always be in gas form, form the lower pressure inside causing it to boil at lower temperatures. The additonal amout of water coul be changing the negative pressure to being more positive then it should be. Edit: There are also so vapor chambers with tall heat pipes coming out of the top them, to help at hotter areas to get the heat in the fins quicker causing the condesing wuicker too. Also the sindering material should be arcross all of the vapor chamber, not just the hot spot of the GPU die. I would shake one those cards up after it was warmed up a little & see if changes anything.

[@matthewandrews2148]

Bottom right is where they pinch off the vapor champers right after its filled with fluid and pressurized, the mesh increases surface area which will allow heat to dissipate, and fluid dynamics and thermal dynamics processes to occur that allow heat to escape. From what I can tell the vapor champers appears to have no design or defects at-least visually. I recommend in your next video Hit the heat sync / vape chamber with a forward-looking thermal imaging systems in your tests, may give you the answer. The one test where you started with it horizontally then turned it vertical wile on was very interesting, and the FLIR may show us exactly what is happening, vs starting vertical and the temp starting and staying lower.

[@SlickR12345]

Or maybe there is no heating issue, but rather a sensor issue that reads the temps incorrectly. I mean come on, what is more likely? Thousands of vapor chambers being designed faulty and no testing catching it, or single sensors being faulty and reading higher temps.

[@FreshJ1v3]

@@SlickR12345 If the temp reading was false then why does the card downclock? Anything is possible but I'm leaning towards a fluid dynamic failure due to the odd behavior when tilted and running then tilted back.

[@flashmhp]

There could still be a problem with the porosity of the material that needs more microscopic views and testing.

[@panemetcircenses510]

One clarification, I think you meant vacuum not pressurized. The lower pressure allows the fluid to turn to vapor at lower temperature.

[@mechitworks]

This was interesting to have a look at and for me to make a quick video about. I'm an engineer that has worked with some heatpipe stuff. I can get behind that there was probably not enough fluid charged. When you don't have enough fluid the working temperature range of a heatpipe or vapour chamber drops. If you go over this ideal working temperature it becomes less effective. If this is the case the idle temperature would also be lower (because the working temperature is lower). you could test this by testing the gpu in a really cold chamber, a lower ambient temperature would allow the vapour chamber to work at a lower temperature as well. It is interesting that rotating the GPU back doesn't drop the temperatures, If it was a classic overcooking event the temperatures should start dropping with more effectiveness. This makes me think it might be an issue with vapour blocking fluid from flowing back or fluid getting "stuck" in certain areas of the Fin stack. This can also be tested by measuring the temperature on the fins. It should have a significantly lower temperature with the gpu horizontal. It is hard to measure if the temperature difference is small.

[@eazen]

At least one useful comment around here.

[@hanp2205]

lol whats a fin stack? sorry idk much about vapour chambers

[@erikhendrickson59]

That vapor chamber looks *_much more expensive_* to manufacture than standard heatpipes.

[@jondonnelly3]

It does, the mind boggles!

[@zwenkwiel816]

so that's where that $1000 price comes from, shame it doesn't work XD

[@Torbjorn.Lindgren]

Well, the comparison here isn't with one heatpipe, if they'd gone that route they'd probably used a whole bunch (8-10?) heatpipes. But yes, even compared to that a big vapor chamber is EXPENSIVE. Which is why smaller vapor chamber which then couples to regular heatpipes are probably more common than this behemoth - and even these smaller rectangular vapor chamber probably costs as much as the heatpipes connected to it. They're also much simpler to design and model than a behemoth like this (and heatpipes are "off the shelf") which further adds to the cost of this one - have to recover the design cost over the manufacturing run.

[@zeus1117]

Yes of course

[@B16B0SS]

AMD did not cheap out on the cooler to keep the card small and then it doens't work properly ... sucks

[@MageThief]

I love stuff like this, when I was a kid I always broke apart stuff to see how it looked inside, to the frustration of my dad 😆

[@Leisur1st]

Same here 😂 he was not happy when I pulled apart the VCR

[@timothysmith160]

@@Leisur1st i regrettably was the same with insects as we did not have computers or VCR's.

[@pamdemonia]

I did it to my friends' toys, much to their dismay!

[@Metalhead-4life]

Prolly cause he was the one paying for the electronics you were twacking apart

[@Apollo-Computers]

I took everything apart too but always put them back together.

[@RageQuitSon]

I would love if companies would show more of these information videos. Maybe like a year after their initial release in case there is any inside secrets. Basically a 'how it's made' straight from the engineers and manufacturers

[@mylarrito]

And it's not like their competitors haven't already done these teardowns a hundred times at launch anyways ^^

[@fpshooterful]

You know whats funny? You can look up how Porsche, Ferrari, Lambo to Royal Royce, etc is made, pretty much from scratch. And these are 100k to 3million+ dollars cars easily. BUT, AMD, or Nvidia can't show the process of how their cards are made? 🤦‍♂

[@PLr1c3r]

If you invested million to billions to get to this point in technology I doubt you would want your roadwork to these innovations revealed for all to see either. This is called intellectual property and by the basic function of capitalism is the core fundamental workings of it.

[@fpshooterful]

@@PLr1c3r I am not saying show the "blue prints" to how the GPU is being made. Heck, all the car Manufacture DOCUs i mention don't even do that. I just want to see the process of how for example the heat chamber are made and installed or just the overall process of the GPU being assembled. At the end of the day, any one can disassemble these cards, as shown here. SO, might as well show how these cards are put together? Again, the same thing they do for these car manufacturers DOCUs. Heck, even Sony DEV teams showed how the PS5 was assembled and dissembled.

[@2009dudeman]

@@PLr1c3r While i'm sure some secrets are kept, cooperate espionage practically ensures there are no secrets for long. Nvidia and AMD may be able to keep wraps on their product while in the design phase, which is critical to maintaining market dominance, but once it hits production there will be a constant flow of leaks from there. The factories have specific arrangements with their respective brand, but that doesn't stop employees or even the company themselves from violating that trust. It would be nice to think that doesn't happen, but it has and will continue. The best example is an event dubbed "The big hack" or the "supermicro motherboard hack". Years back there was a small group of trusted factories making motherboards, these factories had been vetted as trusted suppliers. For one reason or another these trusted factories couldn't meet demand, so they outsourced production without approval to other firms in China. Those firms were not trusted, nor should they have been, as they implanted monitoring chips into electronic devices bound for all levels of the US IT infrastructure. We are talking about spying on the US government without anyone noticing until a few sysAdmins and a few NG firewalls picked up the traffic initially. You can bet secrets are being sold between AMD and Nvidia. The only real limitation is the patents on the design that limits how quickly a competitor can devise the reasoning for a particular design, then modify it such that it falls outside the current patent. At best this buys a manufacturer half a decade, at worst it buys them just long enough for the competitor to retool. Look at the first use of heatpipes in GPUs back in the early 2000s. Nvidia AFAIK pioneered it, AMD came out with their own heat pipe solution for the next model year.

[@watercannonscollaboration2281]

Lots of speculation on what could be going wrong, but it’s always cool watching a heat sink and vapor chamber cut apart to show the insides

[@VoodooZ]

There's a reason lesser youtubers constantly quote you... It's because you're actually do tech journalism as opposed to reporting other people's findings.. Great stuff. Advancing tech, one destroyed cooler at a time! :D

[@txmits507]

You don't need everyone doing the same thing. It takes a lot of time and money to research and diagnosis what he's done. This isn't simple research journalism. Props to him for his knowledge and dedication, but if you need fabrication equipment, it's a specialized situation that you don't casually wander into.

[@27Zangle]

When they quote him, I immediately stop their show and come watch his fully. I feel this is proper considering the effort he puts into his work.

[@VoodooZ]

@@txmits507 Not everything requires equipment though. He's creating content is what i mean. VS reporting news...

[@MaskedMammal]

​@@VoodooZ Not to sound overly mean but I hope most KZbinrs stay in their lane on these matters. It takes a good deal of expertise to properly diagnose issues like this and if they all suddenly decided to "actually do tech journalism" the way we're seeing here, we'll be seeing a huge influx of new misinformation getting around as everyone comes to faulty conclusions. I'm happy to have most of them rounding up relevant tech news and putting it together in digestible and timely formats, and doing consumer-facing reviews on how products are performing across a battery of tests. That is not an insignificant or worthless task in itself.

[@zihechen3111]

@@MaskedMammal those KZbinrs already mislead everyone to think 12900k runs at 280w when 5950x runs at 140w. What can I say? Amd is just better at media control. When intel is the more honest guy and got punished for being honest 😅

[@vanceg4901]

This really helps us visualize what's going on when AMD said the overheating issue was caused by insufficient water in the vapor chamber, thx.

[@haakoflo]

If you have more cards, it would be interesting to see one of them testet horizontally with the fans pointing up.... Edit: That purpose would be to test if the problem is due to the wick drying out (due to not having enough wick between the layers) and to rule out the possibility of other issues related to a horizontal orientation (such as air getting trapped when there is no convection).

[@Quettesh]

He did that in the previous video.

[@haakoflo]

@@Quettesh Are you sure? I scanned through that one again, and didn't find him testing with the fans up, only with the fans down. If he did, maybe you can provide the time spot in that video where he did that test?

[@staples4335]

@@Quettesh No he didn't. He only tested horizontal with fans facing down. (the normal mounting)

[@DS-pk4eh]

@@staples4335 Confirming, he only did standard horizontal orientation.

[@lucidnonsense942]

the upside down orientation you are asking for is inefficient in ALL vapour chambers. You would not be able to say if it's a fault or just the expected inefficiency.

[@GenericPast]

It would be cool to see the manufacturing process of heatsinks like this.

[@jonasduell9953]

Not sure about vapor chambers but iirc Gamer's Nexus went to a big heat pipe manufacturer in China in one of their episodes.

[@marsovac]

attach mesh to the top side, put the supporting pillars, press in the bottom side, weld, from a hole in the side put in a bit of fluid and suck out the air with vacuum, then weld the hole. it is actually much less complicated than it seems. you need to have the tooling though.

[@LeonardoBerrios]

"cool" I see what you did there 😎

[@Alexandra-Rex]

Now that this card has no cooler anymore, it would be cool to see one of Raijintek's big coolers for 120mm fans put on it.

[@haakoflo]

Something tells me EK is already sending a waterblock. Must be a huge marketing potential for them to sell to those who don't want to RMA just for the cooler.

[@TheDeeGeeNL]

@@haakoflo Not everyone can be helped with that. A custom loop is expensive and you need the case and maintenance knowhow for it.

[@sawyerlachance7745]

@@TheDeeGeeNL If you can build a pc you can do water cooling its honestly simpler than building a computer

[@MarvinWestmaas]

@@TheDeeGeeNL I think people paying for an xtx are the easiest audience to upsell a better cooler to, it's not like they are buying the economic option anyway. ... Thinking about it, you're actually right if money wasn't an issue they could have bought Nvidia.

[@DS-pk4eh]

@@haakoflo They are something like 250Euros on pre-order

[@jenda386]

This is awesome. Looks like fabric made of copper. Quite beautiful, really.

[@carnsoaks1]

The mesh is like a tissue, wet it the H2O spreads out, everywhere. So, heating water makes gas, it travels away from the die in theair. It cools and condenses at the edge zones near the fins and fans. The cooling water makes its way back to the die by capillary and the loss / replace physicality of the system . If there is enough fluid, this should work adequately.

[@gertjanvandermeij4265]

LOL ! You are even more retarded than Roman !

[@Sunedosa]

Gotta give you credit, not only did you isolate the issue but you also nailed what caused it : the vapor chamber not having enough fluid. AMD's rep confirming that today is a testament to your work, good job . 👏👏👍

[@jannejohansson3383]

That "fluid" waporize immediately at 20 Celsius and atmosphere pressure. Fluid is not even water, it's some cfc usually. Even mixing many known like30% R134A and 70% R410.

[@johni-db4xv]

Would love to see a comparison with the 4090 vapor chamber. GN had a video with a cross section of the 4090 cooler, but this view of the evaporator side cut out would be very interesting to see side by side. The Nvidia engineer in the GN video mentioned dry out (where evaporation exceeds condensation flow) and how they use a combination of mesh and sintered material in their design.

[@johndoe7270]

Heatsink technology has come a long ways. That looked very tedious. Thank you for taking the time to explore this issue.

[@CallmeRoth]

Yet traditional heatpipe designs are proving to work better.

[@Mehecanogeesir]

@@CallmeRoth Isn't that just companies keeping the cheaper option?

[@Bigboss-xe6lm]

@@Mehecanogeesir To me it looks like a way to save money and call for more money from the customer at the same time...

[@teardowndan5364]

The non-GPU area studs have no wick material likely because there isn't enough heat to bother with it. By only putting sintered material around the GPU's pillars, it probably channels most of the condensate there, where it is actually needed.

[@tessierrr]

Thats what im thinking too

[@spankeyfish]

Also, the material on the GPU pillars doesn't look to have bonded to them which will reduce how much heat they can reject. I think the vapour chamber is too clever for its own good. I bet it works great on paper but it looks difficult to make it well enough to get the performance that it's designed to provide with bonus points for it being sensitive to its orientation.

[@ssaini5028]

@@tessierrr lol sure

[@davidjones6661]

I'd think the surface tension of the water as condensate would be more than enough to pull it to bridge with that height between surfaces, which would allow for position-independent condensate return.

[@teardowndan5364]

@@davidjones6661 The sintered material does double-duty as both a wick to move fluid along and as extra surface area for vapor to condense on. If the sintered material is so over-saturated that liquid can pool up on top and bridge the gap between cold and hot side, you likely lose heat throughput from the sintering not contributing to condensing surface area anymore since it is flooded.

[@smokeyninja9920]

Watched the GN video you mentioned, in it Nvidia's expert (Malcolm Gutenburg) said mesh vs sinter is about porosity, compared sinter to blower coolers (higher pressure, less flow) and mesh to axial fans (higher flow, less pressure) so it sounds like the design is good. They also said they changed the memory contact design to better distribute pressure and get more even contact to the silicon (rechecking your previous video disassembling the xtx it didn't seem to be an issue though). I'm still curious about inverted horizontal performance... Thank you for the time and energy you're spending to help determine what went wrong. I know some people don't understand how valuable this kind of open testing with clear methodology is, and I hope you don't let them discourage you. Even though you can't provide expert analysis of the vapor chamber design, you found a way to open it up and still leave everything basically intact, which opens the door for expert analysis, big kudos for that feat.

[@ThePinkus]

Just a guess, but it might be that the capillary transfer from the cool plate (fin side) to the hot plate is limited to the area in contact with the heat source so to have the liquid phase available where it needs to evaporate, thus taking away the heat from that place by the phase transition. It would be interesting to check if the other places where there are heat sources have the additional material around the support cylinders. Noted that the mesh is present also on the hot plate.

[@jannejohansson3383]

That mesh helps copper parts to weld together when whole pancake is ultrasound welded. There is few parts that they weld with just hear, like capillar connector to put R134A in it.

[@abdulhkeem.alhadhrami]

I love watching this guy tear down defected products to see the insides for a clue to what went wrong, and trying to help the whole community.

[@therealb888]

Happy new year Roman! These 7900XTX series are unparalleled. The detail you go into is much appreciated.

[@rooster1012]

Unparalleled? If you mean subpar, overpriced and defective then I agree.

[@zeus1117]

Yeah it's a complete flop from amd

[@maxdamage4919]

@@rooster1012 LIke melted cables in 4090 ?

[@DeerJerky]

@@zeus1117 lol they're better than a 4080

[@mckagenc080]

@@rooster1012 I'm pretty sure b888 is referring to the series of videos regarding the XTX...

[@haikopaiko]

This is just as amazing as it is interesting! Thanks Roman and the Grizzly team! Again this was very interesting and educational! Thank you!

[@Raintiger88]

Thank you for all the hard work you put into this investigation!

[@skaltura]

nice! :) That's some amazing engineering that goes to vapor chambers, amazing we can build something that intricate at macro scale and at decent cost.

[@lucidnonsense942]

micro not macro ;-)

[@CaptainKenway]

@@lucidnonsense942 He was talking about the mass production aspect of it, so macro would be correct.

[@ausnorman8050]

Okay have watched you're last few vids and this is extremely interesting with the now emerging 7900XTX saga and how it's unfolding. Subscribed!

[@cosminmilitaru9920]

Having opened some heat pipes and vapor chambers myself in the past due to curiosity, without proper tools to do so - only hard labor, it was nice to see yours had some droplets of liquid and moisture, the ones I opened years ago were just very dry inside.

[@SittingDuc]

In addition to the nine "wick" pillars under the gpu hotplate, I expect there are a few "wick" pillars under each dram plate. Would make sense to me. Good teardown, thanks!

[@artemis1825]

Great start to the year! Amazing video

[@DrivenKeys]

Great video, great use of cnc. In GN's Nvidia cooler video, they pointed out two different types of mesh on the flat surfaces of their vapor chamber. In contrast, AMD's is simpler, but larger. I hope they can resolve this without too much loss, and I wonder if horizontal mounting is part of the QC process?

[@TheOriginalFaxon]

One of my friends JUST got a 7900XTX today and is testing it right now, I sent them this to watch! Always love these kinds of destructive teardown videos, there's so much you can learn about a card this way.

[@yakacm]

schadenfreude

[@NewbieTuwbie]

@@yakacm lol

[@erikschiegg68]

Näin käy, kun ei asenna suomalaista saunaa! * *) This is what happens when you don't install a Finnish sauna!

[@randomguy-]

So, did he have the problem, or was it OK? I'm thinking that they have, or at least should have, stopped shipping of the suspected, affected GPU's.

[@yamusa85]

@@randomguy- retailers buy cards in bulk and then sell it to customers. It takes hell to revoke a product quickly. And AMD might be just started tilting its gears toward the problem yet to do anything about it.

[@ChielScape]

Could have measured the refrigerant charge by weighing the cooler before and after puncturing the chamber and boiling out all the liquid. With the dimensional information from the one you opened, the liquid fill % can be calculated at room temperature, and at 110*C.

[@nightshademilkshake1]

this sounds smart but I don't understand it

[@BlackPariah13]

@@nightshademilkshake1 Open GPU, weigh it, boil GPU, weigh it. First weight - second weight = amount of water in vapor chamber (fixed a typo)

[@avetruetocaesar3463]

@@BlackPariah13 *Remove heat sink, weigh it cold, heat it up, weigh it hot.

[@BlackPariah13]

@@avetruetocaesar3463 👍

[@yamusa85]

The problem would be in precision. Cut the nipple open, boil the chamber in a heat oven, measure after 2-3 hours. The problem is it could be chamber pressure fault, where the liquid does not change its states properly on delta although the amount of liquid is correct in ratio to chamber volume.

[@SarcastSempervirens]

Of course! When it's vertical the vapor can move around to different parts but when it's vertical it simply "sits" up the plate like smoke trapped on the ceiling. No movement, no cooling.

[@skerlone]

Most likely the liquid is not getting back or fast enough in the hot gpu area of the chamber because of the shape. Maybe not enough wick material or not good enough in that area. It works well but not always. Even this one worked well in vertical position.

[@timothyandrewnielsen]

Yup. I think the engineers fucked up with that design and thats it.

[@skerlone]

@@timothyandrewnielsen Some combination of the design is too close to the edge of not working combined with manufacturing precision not high enough. Vapor chamber is too big and wide so not enough direct ways for the condensed cooled liquid to wick itself back to the hot side.

[@parthanandi9991]

Looking at you video roman at 3:40 where some of the copper like metal just break apart and seems pretty brittle in nature, and looking at the cloth mesh fibre like structure throughout the vapor chamber e.g., at 7:59, I am wondering if some of this metal bits are too brittle, breaks off and then mixes with the flowing water/liquid thereby clogging the fine fabric like mesh, hampering the free movement of the liquid inside the card ?

[@L0rd_0f_War]

Thank you for the continued testing. Hopefully this will force AMD's hand to actually release some statement about this issue and how they are going to address this moving forward.

[@eazen]

AMD has already acknowledged it, you’re not well informed. They will of course answer later after they are done with their analysis, same as Nvidia with the strange cable.

[@L0rd_0f_War]

@@eazen F Off with your condescending BS. I am well informed about AMDs limited statement a few days ago and an upcoming one (rumoured). I have been following all their statements and news and rumours on every site. I have been running the reddit PSA tech support threads on this 110C issue, and following every piece of news.

[@eazen]

@@L0rd_0f_War the only one arrogant and condescending is you. I corrected you because you’re trying to paint AMD in some negative color, which I rejected. “Force AMDs Hand” as if it ever needed forcing. Guy thinks AMD is Nvidia. Nvidia tried to sell us 4070 for 1000$ dollar as a 4080 branded, they were properly forced to undo that huge mistake. Guy thinks everyone is as shifty as Nvidia.

[@gamehavana120]

@@eazen yea amd is no different than nvidia. amd always can get away from these things because the fanboys are the loudest and toxic.

[@eazen]

@@gamehavana120 yea sure bud. Keep dreaming that up The most toxic fanboys will always be Nvidias and it’s never a contest, if you think otherwise you’ve never been around tech forums

[@konomikitten]

Would it be worth putting a heat source on the GPU area of the vapour chamber then heating it up and pointing a thermal camera at it to see where the hot spots are forming vs the cold spots?

[@chrisfortune1813]

Most likely not in this case as far too much distortion in the removal process although no knowing the internal structure I am sure one could be opened in such a way as to minimise this damage and achieve a meaningful result.

[@Digikidthevoiceofreason]

Vapor*. No U.

[@hrayz]

@@Digikidthevoiceofreason only the US drops the U from words. Canada, UK, Aus, NZ, etc. keep it. Vapour, colour, honour...

[@pvc988]

About 2010 I had orientation dependent cooling problems with Dell 1090 laptop/tablet thingy. Orientation certainly makes a difference for some heatpipe based coolers too.

[@gertjanvandermeij4265]

"Dell" LMFAO !

[@williamdouglass1070]

@@gertjanvandermeij4265chill out

[@zihechen3111]

Heatpipes mostly have less errors due to its mass producing. Unless it’s bent or pressed flat to fit in laptops. Heatpipes on gpu is actually better option on cost wise and error wise. Amd vapor chamber is more like an expensive unnecessary advertisement 😅 those amd incompetences are the reason amd may never be in leading position 😅

[@Smakheed]

The coolant fluid is demineralised water. Your speculation about the solid copper towers is partly correct, they are there to help maintain the chambers height, but also to give direct transfer of heat from the source face to the distribution face of the chamber so that rapid changes of heat can be transferred as fast as possibly to the fin stack for cooling. The mesh is there for the H2o to gain maximum surface area to gather heat, evaporate to the top layer, condensate and where again the maximum surface area created by the mesh layer can dissipate the heat to the transfer surface where it is passed to the fin stack and cooled by the airflow of the fans.

[@GregoryShtevensh]

Under enough pressure, you can increase the boiling point of water by 10° Celsius (boiling at 110°). This is what was done to the He100 prop plane that broke speed records in Germany. And is the main goal of a pressurised cooling system. However, if this was actually water in the vapour chamber, then 110° is on the very limits of "in spec"

[@Tithulta1]

Unreal, I had basic understanding of Vapor chambers, I certainly understand them better now! Thank you! Think I'll go watch that one you mentioned as well!

[@Delistd]

I wonder if the sintered copper powder below the GPU die "pulled away" or "fell away" and prevented the wicking of vapor back to the GPU cold plate?

[@ResidentWeevil2077]

No, what you see in the video is just damage from being scraped with a chisel. Normally the sintered copper would form a solid structure inside the vapour chamber.

[@wecrashgames]

You do an amazing job, Respect to you, Also saw your german channel, Keep on going bro

[@HenrikHvalpen]

Could be interesting to try fill more liquid into one of the faulty ones and see if the fault disappears.

[@Arek_R.]

I'm guessing that once water gets evaporated by the GPU die, the water vapour travels in the vacuum and gets spread out on the surrounding mesh on both sides, and once it turns liquid, there is liquid trapped on separate sides but then it's merged by that sintered material right where the GPU die is and right where the liquid is needed, makes perfect sense. And also on the edges of the vapour chamber both sides are touching right?

[@MarioCRO]

@der8auer just for additional testing, did you try the benchmarks and other testing using AMD platform (AMD 7000 series CPU, with X670E MBO)? Since AMD 7900XTX has some power draw issues depending on monitor make and monitor number, is there a possibility that all these temperatures are due to driver bugs on Intel platform? Probably a long shot, but worth eliminating... For me personally in a closed case with horizontal mounting, my Powercolor 7900XTX works just fine, 3DMark, gaming, Furmark, no issues what so ever, also using 7600X on X670E motherboard.

[@sawyerlachance7745]

The fluid is definatly water this is not a low temp operation and its the most efficent fluid to use. The use of fabric wick material isnt necissarily indicative of low performance it strikes a balance between cost and efficency. If there was a Defect in the fill process water/pressure this seems like a definite possibility however the performance loss on tipping the card doesnt really make sense then. Based on the fact that there is a change when the card is tipped im inclined to believe it is an Interface breakdown between the sintered and mesh sections. If the wick cant return all the fluid back to the cinted section then half of the "Cold side" of vapor chamber wont be returning water back to the "hot side". Not really easy to check without some sort of material defect detection equipment.

[@ys053rious6]

Another excellent video, I love keeping up to date with all the latest stuff with yourself, Jay, gamersNexus and Linus. If it wasn't for you guys within this space I don't think these big companies would ever know there were faults with their products. Its really good to see because they take stock of what you say and hopefully rectify problems. Half the stuff you describe I never fully understand as the technicality goes well over my head, I can build computers and overclock a little but that's it, watching your videos although not fully understanding everything for me is very interesting and enjoy watching all your videos. Heres to another great year for yourself and thank you again for your great content.

[@GregoryShtevensh]

I agree! Although I feel that while J2C is a good channel, Jay just does too much low effort, talking head stuff. Linus tech tips however is probably my favourite channel over all! And Gamers Nexus are brilliant and among the absolute best! I only discovered Roman a couple of days ago, and I feel the tests and experiments he is conducting, put him more on par with GN then just about anyone

[@ys053rious6]

@@GregoryShtevensh J2C was actually recommended to me by grandad on how to sort bent pins on my CPU with the trusty Stanley blade trick, he was the 1st I started watching I do like him as its a bit more chill out and I do value his opinion, Linus is great always full of energy and was my 2nd following I started watching GN for more in depth stuff for the thermals and case stuff and obviously Roman soon followed, I watch kit Guru and RandomGamingHD (more of a random old gen stuff and new stuff benchmark) for a more of a UK perspective and value all of their opinions before buying new tech or builds

[@GregoryShtevensh]

@@ys053rious6 dude I love RandomGaming! Dawid is also awesome! Dawid is funny as. Timmy Joe used to be one of my favs too but unfortunately he ran into some personal issues and stopped

[@ys053rious6]

@@GregoryShtevensh yeah dawid is great with his experiments that he does you never know what he is going to do next all you can pretty much guarantee is that it's going to include vilvida rubber gloves lol

[@crylune]

@@GregoryShtevensh Dawid is pretty fucking obnoxious with his stale innuendoes. And LTT don't know how to do reviews.

[@gotspeed_]

Is it possible to put a water block on the card you took the heat pipes apart from? That card had issues before and it would be interesting to see if the problems are still there after a water block conversion. This would also isolate the problem to the cooler and not anything else. Thank you for the work you do. I have a Z690 board and when your video came out about the issue it helped assure me that my board was good. Again, thanks for all you do.

[@frankderks1150]

To me, because if the hot spots temps higher than 100 degrees, it's more likely that the vacuum wasn't to spec. Another possibility is that the middle stud under the gpu doens't got enough flow because the outer studs under the gpu wicked op most of the fluid. Any change of identifying the hot spots location on the gpu?

[@gertjanvandermeij4265]

LMAO ! If you know shit, than don't try !

[@frankderks1150]

@@gertjanvandermeij4265 Assholes are not preventing me from trying....

[@frostilver]

They used vaper chamber for the die? That's why it has problems. The GPU die area is very hot and after evaporation, the mesh could not soak sufficient water to come back fast enough and do its cooling.

[@3d1e00]

I've always been of the opinion that each heat source needs to be grouped with other heat sources with similar characteristics. You can then size the movement capacity of your cooling to keep what ever those components are at the correct temp. I had a few Nvidia mainboards with huge multi component heat pipe/sink setups. Pretty sure all it did was funnel power delivery heat through the north bridge on way to air. Is a monolithic vapor chamber like that normal? Or do they only use it over GPU and mem and sink power delivery with a plate on other cards?

[@surpriserom]

I just though about it, when you put the card on a vertical mount, any "water" at the bottom of the vapor chamber will be able to go back to the hot side by capillarity as both side will be in the liquid, but if you put the card horizontally, only the pillard in the middle will allow any liquid to go from one side to the other. So maybe the reason it doesn't throttle on vertical mount is the fact that liquid you have the bottom of the cooler help to feed the system liquid back to the hot side, and once vertical, you loose this path and only the pillar in the middle will feed the hot part. Maybe if like you hint, if it miss some liquid in the chamber, it won't have enough liquid to maintain the cooling as the liquid may dry faster than pillard can feed back and if the vertical mount add some way to have more flow back, it may help to keep the temperature in control.

[@Igni-Ferroque]

As if the rings of mesh around GPU core studs are lose. Perhaps when GPU is installed normally they slide and do not have enough contact pressure with backplate side to transfer water efficiently. Amazing video!👍

[@jonasduell9953]

The rings can be lose as long as they touch the nets on both sides they're fine. They just look a bit smashed up from opening, this stuff is super crumbly and brittle because of the extreme porosity it needs for efficient wicking of coolant.

[@Rockport1911]

Normal Heatpipes are already an engineering marvel, but atleast they only have one " street" of heating and cooling going on. It looks like the cooling got lost in this vapor chamber- maze :)

[@FreakyAngelus]

Thank you very much for doing this, I love seeing this and enjoy the reversed engineering steps. As for the actual cooling, those 9 bridges (copper supports) are imo way to thin, weirdly shaped and limited to transport energy away from what you called the GPU heatsink. They work as thermal sources to evaporate the moisture into the chamber through the sintered mesh, which are placed not to mainly transport water from side to side, and transfer energy from heatsink to the other side and into the chamber as evaporated (latent) energy. It's just not a lot of surface area to transfer the amount of heat we're talking about. The water/fluid however does flow back... the extra sintered area around the supports is not placed there for main transport, it goes round on all corners in principle (with the mesh being around the entire circumference) and redistributes on the principle of adhesion (water surface tension... wet cloth will share the fluid). The extra 'connection' with those 9 supports is to create more evaporation area (hence it's raised), transfer energy to the back (hence slightly thicker struts) and evaporate whatever it can (hence the sintered additions) thus quicker dissipate the energy from the heatsink area. The entire principle is evaporative cooling and (while a fallacy to point at my own work with it) something very common to use. Evap chambers are hilariously simple and easy to recreate. The main problem I see in this, is not so much not enough water/fluid, but the long distance for some of it to travel around the block (hence why heatpipes need to get thicker if they get longer). More fluid/water could solve part of the problem. The other element 'wrong' looks the relative simple 9 strut design. It's a very basic cooler approach and with these localized heat spots... not worked on imo. This could be improved upon, a lot... then again, a water cooling block solves most of this as well. The heatsink is not 'thick' imo to spread horizontally enough... and is raised too much... the design is chocking itself. Anywho, thanks for sharing!

[@GSP-76]

Yet there are cards whose junction temps aren't going past 85c when fully loaded. I'd like to know how many cards are defective to have a better idea.

[@FreakyAngelus]

​@@GSP-76 I can speculate, but not give guarantees ;) The junction temperature, as in the previous video, is a good indicator of when it gets overwhelmed and thus runs 'dry'. When you simply have too much heat in the system, then instead of finding an equilibrium between a cold spot (for condensation) and the hot spot (the obvious GPU die on the heatsink), you will run in several issues in the following order. - first you will start to struggle with your hot spots as the fluid can't make it back / evaporate quickly enough from the specific hotspot - second you will fully overwhelm the chamber and all fluid will go into gaseous state. Eventually, just be hot and no longer transferring (internally) the energy - third, the excess of heat can then cause damage/leaks to escape the system and permanently disable the workings. The first is what we see, meaning either the fluid is just on the edge of enough / the routes are on the edge of allowing enough fluid to make it back to keep the cycle going. As tested/demonstrated it seems to do this better in specific positions, but still... seeing those temperatures I would say they are on the edge in all cases. This could be solved with just more fluid.... having a bigger buffer in the chamber and allowing for the prevention of 'running dry'. However... my first glance gives me the feeling the routes are not good/optimized. This means that even with more fluid... if you would continue long enough... it would eventually still run dry (as your system has a choke). Consider it a cycle... fluid evaporates from a hot spot, is a gas... hits a cold spot and condenses. The energy is brought with the medium during that process and absorbed when going into gas state and deposited onto the cold spot when going back into fluid state. If you have a limited amount of fluid, this does not work... but if you have an incorrect surface area to evaporate from... you can also overrun it even while the fluid is trying to flow back. It could be tested... if Der Bauer would drill a tiny little hole in one chamber and add like 1 ml of distilled water... followed by soldering it shut immediately he'd have increased the fluid volume.

[@harrycee656]

The mesh wicks the fluid back to the heat source. The fluid might have evaporated if it has a lower vapor pressure than something like water. Also the cutting process heats up the material further reducing fluid. It may also have only enough fluid to saturate the mesh and sinter copper.

[@poyeep6123]

Hi. Love the way you work man. Question: did you tried to swap heatsink with 1 from properly working not overheating?

[@brandoncorwin8812]

I believe the material around the 9 columns under the processor interface is most likely a ceramic insulator. The insulator is probably there to ensure that the majority of thermal energy generated by the processor is conducted through the columns and into the secondary thermal conduction plate (fin/fan side). Heat is then transferred to the fins and cooled by convection. The vapor chamber aids in conduction of thermal energy from the processor and memory interface plates (exposed copper plates that aren't coated) to the fins on the other side. Buoyancy effects of natural convection transport thermal energy from the hot spots to the rest of the chamber surfaces. This increases the heat transfer rate by transport of thermal energy to other areas of the chamber and out through the fins. The "support columns" you are referring to I believe serve a dual purpose. Sure, they add some rigidity to the assembly. However, they are also conductors of thermal energy from one plate the the other. The problem with the vapor chamber is that it relies on buoyancy forces to transport thermal energy to the other areas of the chamber. Orientation of the card is very important in this regard. However, I believe that these columns you see are meant to account for that should anyone decide to mount these cards on their side. I would be interested to know how many of the reports of overheating are from users that mount the cared on its side to fit it into smaller case or for esthetic reasons. That's my educated guess anyway.

[@nakmail]

Mesh is clearly to increase surface area for condensation. The real question is whether or not you get sufficient cooling area from the mesh alone, or need more from the sintered material. Given that orientation seems to be important in the performance here, it would seem to be down to how the liquid, the heat transport materiel, is flowing around the cooler. Whilst I agree it could be down to a pure volume problem, it could also be a flow problem. The volume of liquid will need to be balanced anyway or you will not maintain low enough pressure to get the right boiling point. So, it’s either design flaw or manufacturing flaw. My guess would be a manufacturing flaw as some cards do not have the problem, but I would also guess this could be too much fluid as well as too little.

[@rajatdixit007]

A slow mo thermal imaging of a good vapor chamber vs suspect vapor chamber might be good. Unsure how much one can see but probably worth a try?

[@victor7491]

I watched this video with my 7900 xtx on my desk. It hasn't been tested yet as I am still waiting for the other necessary parts to arrive, but I wanted to show it what would happen if it decided to be one of those 25.6%

[@Born_Stellar]

over 25% have this problem? woah.

[@victor7491]

@@Born_Stellar yeah it's really bad. Like Roman said, the number is probably inflated but even if it's 10% that's still really really bad for AMD.

[@andresjgm]

Where is exactly the Hot Spot located and where it touch the vapor chamber? What I believe is, this vapor chamber is done just to fight high temps on the GPU, and the vapor/condensation cycle start to fail then more heat is added from other location. If more heat is added, the liquid became vapor again before reach the GPU.

[@s1n1573r-]

Great video as always Roman, Would be awesome to see you cut open a Nvidia 40 series cooler to see more in depth differences in their design.

[@sirius4k]

Gamers Nexus already cut open the 40 series vapor chamber.

[@ALFGamingTV]

beautiful enginering. as every tecnological product it may have some flaws, but just by looking inside and having such huge monstrocity, actually whole cooling solution is a vapor chamber, looks so facinating.

[@zwenkwiel816]

yeah cuz we buy a GPU for the engineering, 110 degrees is fine right? XD

[@ALFGamingTV]

@@zwenkwiel816 i am a forever nvidia user, but it is fascinating to see new products from different companies and cooling solutions. Still, a 100x times better than gigabyte, never ever buy anything from this vendor.

[@pseudonim1]

You just need top find card without that issue and change cooler between cards with problem and without/

[@unpacker9521]

Looks like the copper struts are providing enough heat dissipation such as the core and mcd but there are areas around them that require liquid to cool those areas and that’s where junction temps heat up. It probably would work better if it was either a solid copper plate on top of the gpu with fins coming out or just ditch the vapor chamber and use heat pipes

[@eazen]

Doesn’t explain how the majority of coolers work and only 20-30% are defective. No, the design in general isn’t the problem and can’t be. This is simulated on PCs or tested in person before being used in mass. It can only be a bad batch.

[@timothyandrewnielsen]

Chyna used jizz instead of water

[@adventtrooper]

You could measure the amount of liquid by weighing a sealed heatsink, opening it to allow the vapour to evaporate and weighing again.

[@andreasw.hvammen3946]

Good idea, but very difficult to execute, as the router milling will remove material that then needs to be collected and weighed. Remember we are talking about milliliters (grams) of fluid here. Well, off course you could just pinch a hole, bake it in an oven for a while. That would work actually, if Der Bauer feels like destroying another vapor chamber, that is? Andy;P

[@DailyCorvid]

Freeze it with the moisure inside, then work out with a standard equation the difference between ice and water (in weight). Using that work out the weight of the water when frozen and also liquid... Then you can subtract the water weight accurately. Water when frozen is roughly 1.13 times the weight. So do [weight increase when frozen / 1.13] - [regular weight] = water weight (then to correct it for frozen temp you just divide by 1.13). Something like that.

[@gertjanvandermeij4265]

LMAO !

[@ThatGuy-ht9sp]

@@DailyCorvid But, steel is heavier than feathers

[@thicclink]

@@DailyCorvid please tell me you are kidding lol

[@redsnow846]

So capillary action pulls the liquid through the mesh to the coldplate, kinda like siphoning gas. the heat then causes it to evaporate. the mesh then cools the gas and causes it to condense in the mesh. and the cycle continues. So there are a few possibility's, one would be too little fluid and therefore capillary action ceases to function, there is a crack in the mesh leading to the coldplate stopping capillary action, or the cooler is so overwhelmed the gas is evaporating before it reaches the cold plate.

[@Miracle__Boi]

Thanks for a video! My 7900 xtx works fine, also here in Germany btw, but when this all came up was stressed and run a lot of benchmarks))

[@michahalczuk9071]

Looking at *limited available information,* I can only see that this vapor chamber *seems to be standard,* and it doesn't seem to have any particular problems. You only have those extra wicks on GPU itself, however a lot of liquid will still be able to travel through entire edge from one side to the other. Here is my guess about what exactly is happening here: 1. When card stays *vertical,* the *liquid condensing on cold side can pool on the edge, and be wicked in by the hotter side,* thus giving you enough water for GPU and hotspot to stay low. 2. When card is flipped to *horizontal,* the *liquid pools on the bottom and can only be wicked by those 9 middle struts,* which provide not enough liquid to get even distribution, thus giving very high hotspot temperature, and higher GPU temperature. 3. When you flip the card back, there is *too high temperature differential* between the sides, and the wicking action from whatever is being pooled on the edge cannot *overcome the high condensation rate on cold side* - therefore the hot side runs too dry. Very iconically what happened here is that AMD simply made way too large vapor chamber and/or the edge connection does not allow for good wicking, and also the cooler is too good - it is able too keep too high temperature differential between sides, therefore keeping hot dry. This is very easily solvable though - just add more wicks toward the edge, or closer to GPU. Nvidia didn't run into such problem because their vapor chamber is much smaller and helped with lots of heatpipes. This still means that AMD *should* make a total recall of cards with that design, but I think costs of this action would be at least partially covered by the manufacturer of cooling solution, since it's most likely they designed it.

[@Long-do1vj]

I'm glad you were not offended by our initial comments and suggestions for more investigations. Thank you for looking into it. I really enjoyed watching your content throughout the entire series.

[@der8auer-en]

I'm reading almost all comments on my videos because I appreciate your input and also take it seriously. So if I make a mistake it's just important to know about it :)

[@iNeoN50]

What an extremely deep dive into an issue! Outstanding approach!!!

[@zap117]

can you cool one of the cards with parts laying around, and see if you get any performance boost ?

[@yamusa85]

The question here is not boost but stability.

[@videosuperhighway7655]

Dideuterium monoxide is used for the cooling solution but more expensive than regular Dihydrogen Monoxide. Maybe costcutting is the issue if they tried to use the least amount to save a buck and now you get a recall.

[@vitormoreno1244]

Thank you for the awesome work Roman, about the vapor chamber mesh, Gamers Nexus did a couple of videos on the Nvidia vapor chamber too, check it out when you got the time.

[@Treveliian]

seems to me like a porous surface of that mesh would inhibit the ability to wick the fluid back down the vapor chamber to be evaporated again.. possibly the fluid is remaining trapped in the fibrous pours of the copper material due to static pressure. much like a water droplet stuck in the mesh of a a screen door.

[@st.dietrich437]

It would be cool and intersting to see what is going on in vapor chamber in X-Ray. In motion

[@Stephanthesearcher]

you wont see the movment of vapor and liquid on x.ray

[@Incommensurabilities]

I too would love to see that! However given there's barely any water in them, would an x-ray see anything?

[@st.dietrich437]

@@Incommensurabilities In theory, you could drill a hole and add x-ray contrast substance to liquid inside. Thats how the do x-ray of intestine

[@zerumsum1640]

ok so i have a theory on why mesh was used on the sides, but scintered material was used on the supports: they are controling the heat conductivity of parts so it spreads the heat more. this all has to do with surface area and what each spot is doing. the side facing the cooler's fins is condensing the liquid, and the spots with the scintered pin sleeves are boiling it off. they save money on manufacturing using the mesh on the sides as it's probably easier and cheaper to bond copper mesh to a flat surface than it is to scinter the whole surface. the scintered sleeves over the supports under the gpu cold plate and the rest are done that way because a scintered sleeve is going to be easy to just drop in place over the supports before everything gets sealed up. what i'm not sure about is how they're keeping liquid going to the "hot" parts, it may just be wicking, but i don't know. they're taking advantage of the fact that water boils at a lower temperature at lower pressures, so they can have the state change of the water pull more heat out than if it was just liquid. then, that heat gets dumped into the fins wherever it finds that wall of the cooler. since its at a lower pressure, this has the advantage of letting the expanding steam find its way to the far edges of the cooler a lot faster, spreading all that heat out very quickly, then dumping it and wicking back over to the hot spots. it's a delicate ballence, hopefully they've nailed it.

[@solai]

Great finding! Mistakes happen, even expensive ones... I wonder, what will AMD do now?

[@robotsix6268]

What they should've done at the start: Recall all defective cards. Nvidia is a lot of bad things, but inattentive to their reputation isn't one of them. Trust is earned for decades but dies in mere seconds. Their short-term greed hurts a lot, coming from a closet fanboy. Needless to say, my next GPU is coming from Intel.

[@keldon1137]

Its not a mistake, its basically impossible that this wasnt caught due to how widespread the issue is. They just had to choose between typical amd paper launch or low quality launch.

[@Alex335i]

…and lots of goodwill out the window. They are back to their bench, steps behind Nvidia.

[@georgwarhead2801]

nvidia faces the same problems...all rtx cards had hardware problems at launch 2080ti/3090/4090...and remember, in nvidias case, even if a AIB card does have problems, it is still a problem of nvidias QC since they test every single AIB board and cooler design before they go into mass production. there is not a single AIB card from nvidia wich didnt get tested by nvidia them self and still the QC manage to bring out faulty gpu designes and nvidia also downplayed many of there problems until the public pressure was high enough

[@Hugh_I]

My uneducated wild guess would be that AMD is currently trying to collect serial numbers from people contacting support and crossing their fingers that they can nail it down to one or more bad batches, or vapor chambers from one supplier or something else that allows them to do a limited recall of specific cards, once they are sure they know how to determine which ones are affected.

[@TheZutox]

Do a TEST!! If there is water there I suggest using a syringe, a copper tube and a soldering iron and check the operation of this chamber yourself. Just drill a small hole, solder a piece of tube, then you can add water and suck out some air with a syringe, then clamp and solder the tube.

[@89envision]

Yes! Just cut the tube of the vapor chamber, solder a longer one with valve, then add some water. Then take compressor (from the fridge for example) and suck out some air from vapor chamber. Then close the valve.

[@jolness1]

Can we all appreciate that he does videos in German and English? Glad he does, it’s good content. Also, love that he doesn’t hesitate to say he’s not an expert multiple times.

[@jwo7777777]

The copper is sintered to further increase surface area contact with the fluid and control flow speed through the material. If they are using water, the surface tension of it in liquid form could contribute to lower flow through the sintered or mesh material.

[@Kizmox]

There is easy way to find out how much fluid is in the cooler. Puncture it, measure its weight on high precision scale, then bake it in oven to boil out the liquid and see how much weight changed. You could also compare known good and known bad cooler and find out if there is difference in fluid mass.

[@bullzebub]

the mesh is a vicking material. and yeah. usually its just water in a lower pressure so you get a boiling point around 40 degrees

[@quetzacoatlx]

The water you felt may be the condensed water when the actual low-boiling point liquid evaporated

[@PineyJustice]

The low boiling point fluid used is water. Heatpipes and vapor chambers operate at reduced pressure so boiling starts at a low temperature and as the heat rises the pressure rises bringing the boiling point up so they remain effective in a wider temperature range.

[@haakoflo]

If it's condensed water, it wouldn't work. If it's water, it needs to be very low pressure, or it will not form steam until the GPU temp > 100 degrees C. More likely it's some other liquid with a boiling point higher than room temperature (even in a tropical climate with no AC) but lower than acceptable GPU temps. Somewhere around 50-60C. Edit: As for Wang's comment, I misinterpreted in my response, I think. I first thought that he meant that the "condensed water" was the vapor chamber fluid, and that it was at high pressure (condensed) inside the chamber. I suppose it _could_ be that a very low pressure gas/liquid existed inside the chamber when it was opened, and that it reduced the temperature of the chamber enough to cause condensation when it evaporated. Given the humidity in Germany indoors this time of year, I doubt that, though.

[@PineyJustice]

@@haakoflo It's water at low pressure. As the temperature rises and it boils the pressure increases raising the boiling temperature. This only works because there are condensers dumping the heat keeping it in boiling range. That's also why it's possible to overpower a heatpipe and it will stop working leading to runaway. This isn't a runaway scenario, seems a lot more like the mesh is decoupled on some of these vapor chambers, which would be why it works correctly mounted vertically.

[@haakoflo]

@@PineyJustice A vapor chamber is basically a larger heat pipe, and it can be overloaded in a similar manner. The vapor part is not the bottleneck, though, but the wick part very well may be. The wick creates very significant resistance against the liquid flow, essentially setting a cap on how fast the fluid can flow through. Just as the wick part of a heat pipe. If you have a vapor chamber with the hot part down, you can in principle make it work without a wick, simply by allowing condensed liquid to "rain" back inside the tank. That requires the chamber to be aligned in the opposite direction, though. But even then, you want a wick, since it makes the flow more efficient.

[@PineyJustice]

@@haakoflo Yeah, sorta, the wick doesn't create the resistance you think it does though and we know that these are working in an orientation that isn't raining back down. Fluid wicks up to the hot surface without issue, just it's not wicking from the opposite plate/wick surface. That's why it's working vertically but having issues horizontal, meaning that the wicking posts have possibly de-bonded from the wick of the hot plate.

[@jrodstech]

From a manufacture stand point. It sounds like when making the vapor chamber. The point where it's filled with water failed. We have machines that fill Argon inbetween the panels sometimes doesn't fill properly. But we have test that will tell us it didn't fill and reject the panel. I'm wondering if maybe they had a problem and didn't either realize or they passed them anyway thinking it'll be OK.

[@Cermagine]

Wonder if that discolouration around the middle sintered rod on the gpu die side means anything. Seems a bit unique when compared to the other hot side surfaces.

[@serena-yu]

It seems like a tiny amount of basic copper carbonate i.e. "copper rust", due to copper reacting with CO2 at a hot temperature. It shouldn't matter much unless it gets really thick, which won't happen since CO2 is very limited once the chamber is sealed.

[@bentomo]

This is cool, is it just capillary action that prevents card orientation from being a factor? How does the liquid always get back to the hot spot to change into vapor?

[@fandomkiller]

heat pipes still look superior to vapor chambers. been around longer, keep the laptop space saver heat sinks in laptops please. i would buy an 8 slot card if it was a thing..more pipes lets go. the tooling to make this chamber likely cost a fortune, the copper mesh is insane, the wife's evga 1650 4gb d6 with arctic accelero 4 on it runs 2010mhz core clock (original 1740) +50 mem, 110w power draw, hot spot 56c with 600rpm fan- bigger is better this ting is wild/silent., memory temps not an issue either

[@TheDarkestPhoenix]

If I had to guess, I'd say the lack of sintering around most of the pillars is intentional since a lot of that area is not in direct contact with anything and they're likely just there for structural rigidity. I'd be interested to see if the memory modules also have sintered struts too. I'd guess they would.

[@marsovac]

that's wrong. They are there both for rigidity and as a means for water to drip against gravity through surface/capillary action when the mesh is saturated with condensed water. They put the sintering above the GPU die to help with this process, since that is the point where it most matters. However seeing that there is a breaking point there is not enough fluid for this process to remain sustained so it would hardly matter - if the mesh cannot be saturated then capillary action will not work.

[@gherbent]

I support the Roman's assumption about liquid not traveling back from one side to another due to the insuficient capillary contact. Here is one more effect adding up, the liquid does not travel any more to the area once that reaches certain temperature, similar to floating water drops on a hot frying pan.

[@HamBown]

Unless this is some type of QC issue with manufacturing, I find it pretty hard to believe that there is just not enough fluid for the cooling solution to perform as designed. There were absolutely some highly educated thermal design engineers working on this product and there would have to be some major lack of oversight or testing for this type of problem to make it to mass production. I will be interested to see how AMD responds.

[@allothernamesbutthis]

Sabotage or someone sent the chamber supplyer the wrong drawings?

[@waverleyjournalise5757]

Furthermore, since all of the MBA-edition cards tested by mainstream reviewers functioned almost perfectly with regard to heat dissipation, it's looking more likely that this issue affects certain batches of the cards and *not* all of them. Will be interesting to see how AMD progresses with that.

[@EXG21]

Or they thoroughly tested reviewer cards and not others.

[@waverleyjournalise5757]

@@EXG21 And a test would ensure that the design was not flawed... how? This is what he is trying to test. If the entire design is broken, there wouldn't be anything to cherrypick lol

[@EXG21]

@@waverleyjournalise5757 I meant that AMD made extra sure that reviewer samples were in pristine working condition because of the review process and effects if those cards are faulty.

[@waverleyjournalise5757]

@@EXG21 Yes, but what I mean is that a flawed design is impossible to hide. It's not enough to select "one that works" as if the design is wrong, every card will have the issue. Since some of them do not, it follows that the design itself is not the problem, and that something has gone wrong with the process of how some of them were made. Therefore, the situation will be far easier to solve than if every single card was affected, meaning that dealing with the problem will be far less damaging to AMD. If one factory can be blamed e.g. for manufacturing the faulty vapour chambers incorrectly, then the batch numbers can be traced and the affected cards recalled and repaired. If the cards' design was at fault, then every one of them would show the same behaviour regardless of the mounting pressure, orientation or thermal interface. That would be the absolute worst case scenario.

[@zihechen3111]

@@waverleyjournalise5757 the truth is amd already found the problem before they ship out those cards. However instead of fixing it amd choose to hide it. They send reviewers cherry picked card. Why? Amd runs by shortsighted incompetences at the beginning and just u never realized 😅

[@skywalker1991]

Thanks for your hardworking, happy new year . For now I'm holding off buying rx7900xtx until new revisions come out , its just too risky to pay $1k for faculty product , Hate you amd for making me wait more .

[@thestrykernet]

Sure doesn't look like there are any obvious manufacturing defects which leads me to believe it is something with the evaporating material either not having enough or perhaps wrong mixture.

[@haakoflo]

Not having enough liquid might be an explanation. But i think a more likely explanation is that there is not enough wick to transfer the amount of liquid needed to cool this GPU vertically. Wick is limited to a certain flow (which sets a ceiling on the heat transfer), and it may even dry out completely. If it's turned "inverted horizontally", with fans up, liquid would get help from gravity to get to the other side, and cooling capacity should be increased significantly. If, on the other hand, there is too little liquid, then turning it around will help far less, since there may be too little liquid for the "dew" to form droplets big enough to drip. In that position, a vertical orientation may be the only way (since drops gain size when being pulled by gravity, and then eventually become large enough to drip, as you can see on a car window).

[@InternetEntity]

Interesting... I wonder if the hotspot issue is caused by localised dry-out of the vapour chamber hot area? Liquid coolant soaks back to the GPU contact area where those pillars are. But at the points furthest from each pillar dry out, causing the GPU in direct contact with that bit to receive insufficient cooling? With only the GPU contact area using sintered columns, it is almost like this vapour chamber is one massive condenser with a very small heat soaking area, and there is not enough capacity to return liquid coolant from all around the rest of the chamber to that one small section.