Monday, January 2nd 2023

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

AMD's latest GPUs have been reported to be experiencing overheating issues, with many users claiming that the vapor chamber cooler works better in a vertical rather than a horizontal position. Regardless of orientation, vapor chamber coolers should equal roughly the same heat dissipation performance and move the heat away from the source; however, testing showed that some reference AMD Radeon RX 7900 XTX GPUs feature defect coolers. According to the testing conducted by Roman "der8auer" Hartung, AMD's Radeon RX 7900 XTX RDNA3 GPUs are experiencing problems with overheating caused by a faulty vapor chamber design.

What der8auer found is that these coolers could have a defect in the manufacturing process, where the liquid inside the vapor chamber faces problems in circulation after condensation. It could relate to manufacturing issues of the cooler itself, with an inadequate amount of fluid or insufficient pressure inside the chamber. For more in-depth testing and performance benchmarks, see the video below. It is important to note that we didn't see other reports that replicate this behavior, so always take these reports with a dash of salt.
https://www.techpowerup.com/
Add your own comment

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

#226
Why_Me
oliverprescottI feel like you’re purposely trying to be difficult. Depending the the number of cards affected they might not need to do a mass recall, if they do have to do a recall they’ll probably inform the people specifically affected. There’s also the chance that while only a specific batch was affected not all the cards in that particular batch have the defect. Meaning that doing a recall of the entire batch would be over kill as you could be making people return cards that have no issue. Again some details matter.
If you send out a faulty product and you admit it's faulty you recall said product and replace it with one that isn't screwed up or you refund the customer. Anything less is shady business practice.
Posted on Reply
#227
maxfly
Why_MeIf you send out a faulty product and you admit it's faulty you recall said product and replace it with one that isn't screwed up or you refund the customer. Anything less is shady business practice.
So by your logic, had they simply pulled an ngreedia and ignored the issue, everything would be copacetic...gotcha.
Posted on Reply
#228
oliverprescott
D
Why_MeIf you send out a faulty product and you admit it's faulty you recall said product and replace it with one that isn't screwed up or you refund the customer. Anything less is shady business practice.
Did you even read what you quoted? Also if you’d been paying attention you’d understand the entire product line isn’t faulty. Just specific ones have the defect. Any company that did a full recall of a product line when they knew only a limited number had the issue would be a pretty stupid company.

and there are options for replacing and refunding customers without doing a recall. Seriously either you’re purposely being dense or you’re trolling.
Posted on Reply
#229
Why_Me
maxflySo by your logic, had they simply pulled an ngreedia and ignored the issue, everything would be copacetic...gotcha.
You mean people too ignorant to make sure their cables are plugged in correctly and yes there were less than fifty reported cases. Here's a pro tip. Make sure your case is large enough to fit said card, use dielectric grease on the male ends, be an alpha male and use that testosterone to plug in the connector all the way making sure you have a tight fit.
oliverprescottD

Did you even read what you quoted? Also if you’d been paying attention you’d understand the entire product line isn’t faulty. Just specific ones have the defect. Any company that did a full recall of a product line when they knew only a limited number had the issue would be a pretty stupid company.

and there are options for replacing and refunding customers without doing a recall. Seriously either you’re purposely being dense or you’re trolling.
Get back to me when AMD recalls those defective cards. Anything less is shady business practice.
Posted on Reply
#230
nguyen
come summer and people with seemingly good cooler find their 7900XTX also overheat :rolleyes:.

My guess is there's a thermal threshold that once reached, the vapor chambers that were not properly filled will become ineffective, running the fans at high RPM constantly and low ambient temp can prevent vapor chamber from drying out.

Very solid gaming experience to run your 1000usd+ GPU at deafening noise level, 10/10 product.
Posted on Reply
#231
Why_Me
What I've learned over the years about the AMD cult. They love bugs and they love tweaking their system including ghetto work-a-rounds. That and they'll take one for the team when it comes to defending AMD products.
Posted on Reply
#232
maxfly
Why_MeYou mean people too ignorant to make sure their cables are plugged in correctly and yes there were less than fifty reported cases. Here's a pro tip. Make sure your case is large enough to fit said card, use dielectric grease on the male ends, be an alpha male and use that testosterone to plug in the connector all the way making sure you have a tight fit.
You said it not me.
There's a reason their called nvidiots btw.
You forgot a few steps...just sayin.
Posted on Reply
#233
AusWolf
Why_MeGet back to me when AMD recalls those defective cards. Anything less is shady business practice.
How would they recall the bad cards if they don't know which cards are bad? That's why they said it in the statement: if you have a bad card, contact AMD support, and they'll take care of it. What more do you want?
oliverprescottOf course he’s doing damage control after being wrong about it, he’s just doubling down on his original mistake. I’ve honestly lost a lot of respect for him over this.

And now with his whole gpu going up in smoke I don’t know what he’s trying to prove. He goes on and on about understanding that these two issues probably aren’t connected and talks about not making bad PR for AMD but why even have that in the video if not to conflate the two things? He says one thing but his actions are the opposite. Like what purpose does this video serve other than propping himself up and trying to make amd look bad even though they’ve made a reasonable response to the whole thing?

If anything this video further proves that buying known broken cards from your own viewers to try and prove that the issue is widespread is not a good way to do that. There’s clearly something wrong with that card that doesn’t have anything to do with the vapor chamber and that in itself could be causing the performance and heating issues. And on top of that we don’t actually know what was done to that card before he got it. There’s just too many holes in his testing and reasoning.

And just wait and see there will be a ton of other YouTubers digging in along with him saying now the over heating issue is starting fires in gpus. Regardless of him saying not to conflate the two he’s clearly linking them together by making this video. That’s some next level double speak he’s got going on.
He's obviously trying to defend the poor innocent consumers from evil corporations like AMD. Because overpaying for Ngreedia is so much better! Errors on that side just do (link) not (link) happen (link).

It's totally not about the views. At all. :rolleyes:
Posted on Reply
#234
Why_Me
AusWolfHow would they recall the bad cards if they don't know which cards are bad? That's why they said it in the statement: if you have a bad card, contact AMD support, and they'll take care of it. What more do you want?


He's obviously trying to defend the poor innocent consumers from evil corporations like AMD. Because overpaying for Ngreedia is so much better! Errors on that side just do (link) not (link) happen (link).

It's totally not about the views. At all. :rolleyes:
Like Debauer pointed out, if it's a small batch of cards like AMD claims they will have the serial / batch numbers.
Posted on Reply
#235
AusWolf
Why_MeLike Debauer pointed out, if it's a small batch of cards like AMD claims they will have the serial / batch numbers.
Because they obviously know which batch they're all from without anyone ever RMA-ing any of them.
Posted on Reply
#237
Chomiq
P4-630Amd/comments/106h3z0

Magic smoke edition.

Looks like completely separate issue, derBaur needs to send the board to Buildzoid.
Posted on Reply
#238
kapone32
Why_MeLike Debauer pointed out, if it's a small batch of cards like AMD claims they will have the serial / batch numbers.
How many countries were selling the cards, How many provinces were the cards sold in, how many Cities were the cards sold in, How many retailers were the cards sold in? How many cards were sold online.
Posted on Reply
#240
AusWolf
Why_MeAsk AMD how many of those faulty cards were manufactured because they know.

www.guru3d.com/news-story/amd-says-it-is-out-of-stock-to-replace-faulty-radeon-rx-7900-xtx-gpus.html
How did you come to that conclusion? The article says they're out of stock to replace the faulty ones... which is in no connection to what you're saying whatsoever.

The bad ones need to come back en masse first to see which batch(es) are affected. If you made 1000 GPUs, and let's say 20 of them are faulty, how do you know which 20 unless people RMA them?
Posted on Reply
#241
Why_Me
AusWolfHow did you come to that conclusion? The article says they're out of stock to replace the faulty ones... which is in no connection to what you're saying whatsoever.

The bad ones need to come back en masse first to see which batch(es) are affected. If you made 1000 GPUs, and let's say 20 of them are faulty, how do you know which 20 unless people RMA them?
There's a lot of messed up cards and they know this. It's a design issue.
Posted on Reply
#242
kapone32
Why_MeAsk AMD how many of those faulty cards were manufactured because they know.

www.guru3d.com/news-story/amd-says-it-is-out-of-stock-to-replace-faulty-radeon-rx-7900-xtx-gpus.html
Do you really believe that every single person that bought a 7900XTX registered their card? The people that have contacted them for RMA are the ones that they are aware of. That does not mean that the ones returned to the retail store were included in that tally however. Logistics is good but not that good. Some people only install the driver and don't even worry about the software and as I said before that is not as easy as you think for AMD to contact all the people that have sub optimal cards because they know who they are.
Why_MeThere's a lot of messed up cards and they know this. It's a design issue.
Too little coolant is a design issue? Sounds like manufacturing to me. If every single card was compromised they would not be avialable to buy retail.
Posted on Reply
#243
OfficerTux
Igor has published a new in-depth analysis of the supply chain and quality control mechanisms regarding the issues with the vapor chamber.

Unfortunately it is only available in German: www.igorslab.de/einzelfaelle-oder-vapor-gate-samt-versagen-des-qualitaetmanagements-hintergruende-zum-rx-7900-xtx-kuehler-problem-samt-qm-qa-und-qc/

Here is a short rather long summary:
  • The whole card is manufactured by PC Partner, the cooler is made by Cooler Master, the vapor chamber is supplied by AVC (Asia Vital Components).
  • As we all know by now the issue is caused by too little liquid in the vapor chamber. As far as Igor knows one charge of vapor chambers from AVC was affected which usually consists of 10000 units.
  • Not all vapor chambers from that charge are fully defective, some work ok-ish, some show the 110°C issue.
  • This slipped through quality control at Cooler Master and PC Partner which casts a damning light on their quality processes.
  • To make matters worse vapor chambers and coolers are not serial number matched. That means the defective charge of 10000 vapor chambers is split among several charges of coolers and was mixed with non-defective charges.
  • This means AMD has no way to match the serial numbers of the cards with the defective coolers, they can not issue a list of defective cards and the customer has to find out by himself.
  • The amount of cards affected is quite high, AIBs (like Saphhire and Power Color) report ~10% of RMAs of their cards.
  • At the beginning AMD support offered replacements within 2 weeks, now they say they can not provide a delivery date as their shelves are empty and rather propose a refund. This might be worsened by the problem that they themselves do not know which cards are affected and do not want to ship out another faulty card to an already unsatisfied customer.
  • For us end consumers this is annoying, for the bigger system integrators who build hundreds of systems this is a real problem since they now sit on unfinished PCs they can not sell.
Posted on Reply
#244
Dirt Chip
What a headache for system builders, hope AMD will take care of them too.
At 10% RMA it's no small problem and thousands of cases sounds reasonable.

At least it's just pref hit, no real danger so even if affected you can keep use it albeit at the cost of noise. When summer come it will only get worse.
Posted on Reply
#245
kapone32
Dirt ChipWhat a headache for system builders, hope AMD will take care of them too.
At 10% RMA it's no small problem and thousands of cases sounds reasonable.

At least it's just pref hit, no real so even if affected you can keep use it albeit at the cost of noise. When summer come it will only get worse.
In some ways this could be good. We could get from AMD Heavily discounted Refurbished units. The would not last long but if they sold them refurbished on their website they would but they will probably drop them rightr back into supply chain. That Youtube video where the producer uses a syringe to fill the Vapor chamber is actually sublime.
Posted on Reply
#246
AusWolf
OfficerTuxIgor has published a new in-depth analysis of the supply chain and quality control mechanisms regarding the issues with the vapor chamber.

Unfortunately it is only available in German: www.igorslab.de/einzelfaelle-oder-vapor-gate-samt-versagen-des-qualitaetmanagements-hintergruende-zum-rx-7900-xtx-kuehler-problem-samt-qm-qa-und-qc/

Here is a short rather long summary:
  • The whole card is manufactured by PC Partner, the cooler is made by Cooler Master, the vapor chamber is supplied by AVC (Asia Vital Components).
  • As we all know by now the issue is caused by too little liquid in the vapor chamber. As far as Igor knows one charge of vapor chambers from AVC was affected which usually consists of 10000 units.
  • Not all vapor chambers from that charge are fully defective, some work ok-ish, some show the 110°C issue.
  • This slipped through quality control at Cooler Master and PC Partner which casts a damning light on their quality processes.
  • To make matters worse vapor chambers and coolers are not serial number matched. That means the defective charge of 10000 vapor chambers is split among several charges of coolers and was mixed with non-defective charges.
  • This means AMD has no way to match the serial numbers of the cards with the defective coolers, they can not issue a list of defective cards and the customer has to find out by himself.
  • The amount of cards affected is quite high, AIBs (like Saphhire and Power Color) report ~10% of RMAs of their cards.
  • At the beginning AMD support offered replacements within 2 weeks, now they say they can not provide a delivery date as their shelves are empty and rather propose a refund. This might be worsened by the problem that they themselves do not know which cards are affected and do not want to ship out another faulty card to an already unsatisfied customer.
  • For us end consumers this is annoying, for the bigger system integrators who build hundreds of systems this is a real problem since they now sit on unfinished PCs they can not sell.
This is probably the most useful post along the 10 pages in the entire thread. It should be pinned to the top.

I wonder where der8auer is now to tell us that it's all crap and AMD hates everyone.
Posted on Reply
#247
nguyen
Man AMD is an endless source for meme 2023
Posted on Reply
#249
kapone32
OfficerTuxIgor has published a new in-depth analysis of the supply chain and quality control mechanisms regarding the issues with the vapor chamber.

Unfortunately it is only available in German: www.igorslab.de/einzelfaelle-oder-vapor-gate-samt-versagen-des-qualitaetmanagements-hintergruende-zum-rx-7900-xtx-kuehler-problem-samt-qm-qa-und-qc/

Here is a short rather long summary:
  • The whole card is manufactured by PC Partner, the cooler is made by Cooler Master, the vapor chamber is supplied by AVC (Asia Vital Components).
  • As we all know by now the issue is caused by too little liquid in the vapor chamber. As far as Igor knows one charge of vapor chambers from AVC was affected which usually consists of 10000 units.
  • Not all vapor chambers from that charge are fully defective, some work ok-ish, some show the 110°C issue.
  • This slipped through quality control at Cooler Master and PC Partner which casts a damning light on their quality processes.
  • To make matters worse vapor chambers and coolers are not serial number matched. That means the defective charge of 10000 vapor chambers is split among several charges of coolers and was mixed with non-defective charges.
  • This means AMD has no way to match the serial numbers of the cards with the defective coolers, they can not issue a list of defective cards and the customer has to find out by himself.
  • The amount of cards affected is quite high, AIBs (like Saphhire and Power Color) report ~10% of RMAs of their cards.
  • At the beginning AMD support offered replacements within 2 weeks, now they say they can not provide a delivery date as their shelves are empty and rather propose a refund. This might be worsened by the problem that they themselves do not know which cards are affected and do not want to ship out another faulty card to an already unsatisfied customer.
  • For us end consumers this is annoying, for the bigger system integrators who build hundreds of systems this is a real problem since they now sit on unfinished PCs they can not sell.
I surmise the first batch of reference cards is usually the only mass produced run. Maybe there now will be another run though. At least they caught it quickly and AMD are working hard to fix it. When you talk about QC there is help needed everywhere. I went to Niagara falls late last year and in my hotel they had 1 person working at the Starbucks downstairs.....in a Vacation town on a Saturday morning? This is not good but at least it is much clearer now. Thanks for posting.
OfficerTuxThat is not totally true. In the cases where the hotspot hits 110°C the DRAM often also is at > 100°C and it is not rated for those temps so it might shorten its lifespan.

(Example: www.digikey.com/en/products/detail/micron-technology-inc/MT61K512M32KPA-16-B/10321310, rated up to 95°C)
My memory sits at 58C for Memory Junction temp at idle in HWinfo64. I had a reset and zero fan was reapplied I have seen that temp as high as 96 C.
Posted on Reply
#250
OfficerTux
AusWolfThis is probably the most useful post along the 10 pages in the entire thread. It should be pinned to the top.

I wonder where der8auer is now to tell us that it's all crap and AMD hates everyone.
First of all, thanks for the kudos.

But I feel like many here did not see the full videos from der8auer but just pick the parts that make him look bad. For an example in his follow up video he explicitly says that the issue with the broken card he checks may be related to the vapor chamber or not and that you have to be careful with single cases and that they do not reflect the majority of cards.

I like his videos, but that's just my opinion. I watch them in German though, maybe his English translations are not as good.
Posted on Reply
Add your own comment
Jun 15th, 2024 18:13 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts