AMD Radeon RX 7900 XTX May Feature Faulty Coolers, Causing Overheating

[zlobby]

AMD is not making coolers. They design them, the manufacturing is not done by them.

Yet some of their (AMD's) cards allegedly have bad(ly performing) coolers.

And of course AMD is not manufacturing coolers. Heck, they stopped making the chips themselves a long time ago.

I'd argue even that AMD could be paying for someone to design the coolers for them, just as they do with chipsets.

 

[Dirt Chip]

Thanks for the summary, saved a watch.

Also he show a survey by "ComputerBase" in german, so far ~25% of 223 vote they have this issue.
It is for sure biased towards higher rate then true, but show the general scope (that is not sub 1% as with 12vhpwr).

 

[Vya Domus]

Yet some of their (AMD's) cards allegedly have bad(ly performing) coolers.

And of course AMD is not manufacturing coolers. Heck, they stopped making the chips themselves a long time ago.

I'd argue even that AMD could be paying for someone to design the coolers for them, just as they do with chipsets.

I don't know if you understood what I said, AMD does not make coolers they just design them. And the coolers that they design work fine, look at the TPUs review of the 7900XTX where it reaches only 73C hotspot temperature. This is in an open test bench I think, nonetheless, there is clearly nothing wrong with the design. Why would they pay someone else for something that works just fine ? Maybe this was a QA issue that could have been spotted by AMD, I don't know how that process works, although it is still not clear to me how many of these things are faulty or if that's even the issue here. Nonetheless, if that is the problem, sadly AMD might have not had the ability to prevent this since they are not in charge of the manufacturing.

so far ~25% of 223 vote they have this issue.

223 is a very small sample size, not the mention the obvious bias in being more likely to answer the pool if you do in fact have an issue. It's also not clear what "the issue" even is, do all of those people have cards that hit 110C ? Or do they hit other, lower, temperatures that the users find problematic.

 

[TheoneandonlyMrK]

AMD 7000 Series Recall - RX 7900 XTX & XT:

And this is the shit that comes of such hyperbolic, conclusions, regurgitated Chinese whisper style, next go round alll AMD cards need recall.


Der8aure with millions of views comes a modicum of responsibility.

That guys dropping in my estimation I'll watch this new video of his.

 

[nguyen]

So far I have seen 4 long time members of Guru3d whom have high hot spot temp on their 7900XTX MBA and already refunded them (got 4080/4090 instead).

AMD just keep making sure enthusiasts buy Nvidia instead

 

[ZoneDymo]

AMD 7000 Series Recall - RX 7900 XTX & XT:

And this is the shit that comes of such hyperbolic, conclusions, regurgitated Chinese whisper style, next go round alll AMD cards need recall.
Der8aure with millions of views comes a modicum of responsibility.
That guys dropping in my estimation I'll watch this new video of his.

If you would tell me the guy in the vid is a bot, I would believe you, dear lord what a generic voice, delivery, setup....not an ounce of personality....and yeah then the sensationalist bs as well....yikez.

 

[Zyll Goliat]

I personally do not own RX 7900XTX but I recently get myself RX 5700 and because that card have reference cooler and I hate the noise I replace the cooler from some old ATI 4890 card....now I am saying this because while I was doing all of this I played a lot trying to mount and test everything as it should be and In that process I notice that sometimes the problem appear with the junction temperature even if the gpu temps staying bellow 65c but junction temp gets more then 20c higher and then card could become unstable.....I used the fine sandpaper to polish the surface of the cooler and at end everything working well but this led me to believe that maybe in this case possibly GPU die or the cooler surface is not perfectly flat and that maybe could be some tiny micro pockets where the higher temps accumulate and that's why when you move the card in the certain position the temps could get better as the gravity+termal paste will fill that pockets better and spread the temps.....saying this as just as my theory crafting and nothing more.........

 

[ZoneDymo]

Yet some of their (AMD's) cards allegedly have bad(ly performing) coolers.

Alleged when and by whom?
AMD has perfectly fine adiquate coolers since forever, want something better with some nice OC headroom? buy a partner model or go watercooling or so.
That is the relationship between AMD and boardpartners.

Nvidia has actually screwed them over with their latest releases, extremely competent (overbuild even) reference coolers and an early release for Nvidia exclusively, making people buy directly from Nvidia with no need for a partner model.
Hence EVGA is out and as was suggested, Nvidia is looking more towards being like Apple, full and exclusive control of the entire manufacturing chain.

I personally do not own RX 7900XTX but I recently get myself RX 5700 and because that card have reference cooler and I hate the noise I replace the cooler from some old ATI 4890 card....now I am saying this because while I was doing all of this I played a lot trying to mount and test everything as it should be and In that process I notice that sometimes the problem appear with the junction temperature even if the gpu temps staying bellow 65c but junction temp gets more then 20c higher and then card could become unstable.....I used the fine sandpaper to polish the surface of the cooler and at end everything working well but this led me to believe that maybe in this case possibly GPU die is not perfectly flat and that maybe could be some micro difference on the surface of the die where the higher temps accumulate.....saying this as just as my theory crafting and nothing more.........

Well you are absolutely correct, extremists used to lap dies as well, though its again, a pretty darn extreme measure, it is what thermalpaste is for, bridging those little imperfections.
Also I desire a picture if you have some of your handywork

 

[AusWolf]

Ok here is the latest de8auer video......

TL;DR
Structure seems fine, might be not enough H2O inside.
Also he show a survey by "ComputerBase" in German, so far ~25% of 223 vote they have this issue.

TL,DR 2: We didn't know if the vapor chamber was at fault in the first place, but now that he opened one up, we still don't know. Oh, and people with problems voted that they have problems. Very useful again.

(before anyone jumps on me again, I watched the full video this time)

By the way, where's Gamer's Nexus when we need them? (Maybe trying to make a PROPER analysis?)

 

[Bomby569]

TL,DR 2: We didn't know if the vapor chamber was at fault in the first place, but now that he opened one up, we still don't know. Oh, and people with problems voted that they have problems. Very useful again.

(before anyone jumps on me again, I watched the full video this time)

By the way, where's Gamer's Nexus when we need them? (Maybe making a PROPER analysis?)

unless there is some curse on the cards, he pretty much eliminated everything else and if there isn't any leprechaun inside that doesn't like to go horizontal, it can only be the vapour chamber.

 

[Dirt Chip]

I don't know if you understood what I said, AMD does not make coolers they just design them. And the coolers that they design work fine, look at the TPUs review of the 7900XTX where it reaches only 73C hotspot temperature. This is in an open test bench I think, nonetheless, there is clearly nothing wrong with the design. Why would they pay someone else for something that works just fine ? Maybe this was a QA issue that could have been spotted by AMD, I don't know how that process works, although it is still not clear to me how many of these things are faulty or if that's even the issue here. Nonetheless, if that is the problem, sadly AMD might have not had the ability to prevent this since they are not in charge of the manufacturing.

View attachment 277307


223 is a very small sample size, not the mention the obvious bias in being more likely to answer the pool if you do in fact have an issue. It's also not clear what "the issue" even is, do all of those people have cards that hit 110C ? Or do they hit other, lower, temperatures that the users find problematic.

It's not AMD making the cooler, but it's their name on it so they responsible for any problem it might have. It is their reference card after all. This also need to be said.

And you choose to quote just one line from my comment about the survey, deleting what I wrote next, just to write exactly the same thing yourself, about the biased sample. I find it strange.

 

[TumbleGeorge]

Hmm, height(and volume) of steam chamber may is not enough for properly liquid-gas phases?

 

[AusWolf]

unless there is some curse on the cards, he pretty much eliminated everything else and if there isn't any leprechaun inside that doesn't like to go horizontal, it can only be the vapour chamber.

He said the problem remained when he flipped the cards back, didn't he? Maybe it isn't related to orientation - it's just a coincidence? (not conclusion, just speculation)

Edit: Could there be a piece of BIOS or driver code that stresses one part of the GPU more than others? Maybe it's a far-fetched idea, but I'm not a programmer, so it doesn't hurt to ask, I guess.

 

[Bomby569]

He said the problem remained when he flipped the cards back, didn't he? Maybe it isn't related to orientation - it's just a coincidence? (not conclusion, just speculation)

Edit: Could there be a piece of BIOS or driver code that stresses one part of the GPU more than others? Maybe it's a far-fetched idea, but I'm not a programmer, so it doesn't hurt to ask, I guess.

when you flip the heated vapour chamber you can't expect it to behave as cold.

Come on man BIOS or any software issue, or even the fans, the heatsink itself, none of those things care if the card is vertical, horizontal or any other position while outside of a case. Only the vapour chamber makes sense.

 

[AusWolf]

when you flip the heated vapour chamber you can't expect it to behave as cold.

Come on man BIOS or any software issue, or even the fans, the heatsink itself, none of those things care if the card is vertical, horizontal or any other position while outside of a case. Only the vapour chamber makes sense.

So if you experience the problem, turn off the PC and let the card cool down before flipping it back into its original position, the problem goes away? He isn't very clear on this (which he should be if he suspects the vapor chamber is the issue).

 

[Bomby569]

So if you experience the problem, turn off the PC and let the card cool down before flipping it back into its original position, the problem goes away? He isn't very clear on this (which he should be if he suspects the vapor chamber is the issue).

it obviously doesn't solve any problem, any of the problematic cards have this issue every time they heat up when in the horizontal, as per the user reports. The only solution is not using it or mounting vertically and many people simply can't do it, small cases or cases that don't allow it.

 

[AusWolf]

it obviously doesn't solve any problem, any of the problematic cards have this issue every time they heat up when in the horizontal, as per the user reports. The only solution is not using it or mounting vertically and many people simply can't do it, small cases or cases that don't allow it.

But the question is... does the problem go away every time and with every card when you orient them vertically?

My point is that if there's a solution already, then it's easier to identify the problem.

 

[Dirt Chip]

TL,DR 2: We didn't know if the vapor chamber was at fault in the first place, but now that he opened one up, we still don't know. Oh, and people with problems voted that they have problems. Very useful again.

This is funny sarcasm, not more.
As with 12vhpwr, we had many fallow up videos even from GN until a very conclusive finding was to shown.
I welcome those follow up as it allow better understanding (even if slightly) and It might be of value later on.
TPU also have the same survey with about the same result by now (30-70, n=105).
If you are not interested in the process discovering what`s with, you can just avoid it completely.

 

[Bomby569]

But the question is... does the problem go away every time and with every card when you orient them vertically?

yes, according to Derbauer. But is that any solution? NO.

 

[AusWolf]

This is funny sarcasm, not more.
As with 12vhpwr, we had many fallow up videos even from GN until a very conclusive finding was to shown.
I welcome those follow up as it allow better understanding (even if slightly) and It might be of value later on.
TPU also have the same survey with about the same result by now (30-70, n=105).
If you are not interested in the process discovering what`s with, you can just avoid it completely.

Such surveys are always flawed because people with problems are a lot more inclined to vote. People without problems just get on with their lives. It's kind of the same as product reviews. Reading them is fine, but when you compare numbers, you should always take a huge pinch (or more like a kg) of salt. People with extreme negative experiences are a lot more likely to voice their opinions.

As for follow-ups:

yes, according to Derbauer. But is that any solution? NO.

It's not a solution, but a clue to it.

That's what I wanted to know - if he actually tested this.

 

[Why_Me]

Luckily for AMD they have a decent share of the server market to fall back on in case they end up closing down their gpu division.

 

[Aquinus]

yes, according to Derbauer. But is that any solution? NO.

Kind of. Once the vapor chamber is heat saturated and dried out, there is no going back. My theory is that reducing the power limit will probably solve the issue for those experiencing it. I believe that the vapor chamber is near its thermal limits. A lot of vapor chambers are designed to work under 100°C. As soon as the vapor chamber dries out, thermal conductivity decreases significantly which further exacerbates the inability for the vapor to re-condense because the the chamber will be hotter now due to the change in thermal conductivity. All in all, I think they just pushed the vapor chamber too hard and there are certain situations that will exacerbate this issue, such as the orientation of the card. I think this is the case because of how Derbauer changed the orientation back, but temps still remained crappy.

 

[nguyen]

Time for AMD to release the 7900FIXed edition, for only 1099usd

 

[eidairaman1]

They need to fire some people asap, for years of work to be wasted just because someone messed up the cooler is unacceptable.

How about making them fix the design if there is a flaw instead?

 

[AusWolf]

Kind of. Once the vapor chamber is heat saturated and dried out, there is no going back. My theory is that reducing the power limit will probably solve the issue for those experiencing it. I believe that the vapor chamber is near its thermal limits. A lot of vapor chambers are designed to work under 100°C. As soon as the vapor chamber dries out, thermal conductivity decreases significantly which further exacerbates the inability for the vapor to re-condense because the the chamber will be hotter now due to the change in thermal conductivity. All in all, I think they just pushed the vapor chamber too hard and there are certain situations that will exacerbate this issue, such as the orientation of the card. I think this is the case because of how Derbauer changed the orientation back, but temps still remained crappy.

This might be a stupid question, but how does a vapor chamber dry out? It's a closed system, which means vapor can't escape, so it can't dry out, can it?