wccftech AMD’s Confirms Radeon RX 7900 XTX Thermal Issues Related To Vapor Chamber Coolant, Offering RMA To Affected

[bonami2]

Quote

The interview was done with AMD's Senior Vice President & General Manager of the Graphics Unit, Scott Herkleman who confirms that the Radeon RX 7900 XTX reference graphics cards indeed have a faulty vapor chamber on a small batch of cards that weren't filled with enough coolant and is the prime cause of the overheating issues that causes GPUs to run at up to 110C Junction temps and cause thermal throttling.

AMD's Graphics SVP Confirms Radeon RX 7900 XTX Thermal Issues Related To Vapor Chamber Coolant, Offering RMA To Affected

WCCFTECH.COM

---

AMD finally confirms that the Radeon RX 7900 XTX reference graphics card has a faulty vapor chamber that causes overheating & will offer...

 

[Mr. Fox]

This has been a really bad couple of years for GPUs.

• Crypto-scalping
• Post-crypto scalping
• Supply/demand manipulation
• Unlaunched inferior version of the 4080
• NVIDIA 12+4 pin connectors burning
• Misrepresentations of 7900 XT/XTX performance
• Continued manipulation/suppression of 4090/7900 XTX supply
• Relaunched 4070 Ti (the inferior 4080) grossly overpriced
• Defective AMD vapor chambers
• AMD drivers killing last-gen GPUs

 

I likely forgot about something that should be on the list.

 

Edited January 11 by Mr. Fox

[pioneerisloud]

1 hour ago, Mr. Fox said:

This has been a really bad couple of years for GPUs.

• Crypto-scalping
• Post-crypto scalping
• Supply/demand manipulation
• Unlaunched inferior version of the 4080
• NVIDIA 12+4 pin connectors burning
• Misrepresentations of 7900 XT/XTX performance
• Continued manipulation/suppression of 4090/7900 XTX supply
• Relaunched 4070 Ti (the inferior 4080) grossly overpriced
• Defective AMD vapor chambers
• AMD drivers killing last-gen GPUs

 

I likely forgot about something that should be on the list.

 

Would you mind expanding on that last point a little?  That's a first I'm hearing about that one, and I have a lot of AMD GPU's I don't want ded.  Unless you're just meaning the drivers that absolutely nuke performance to the point last gen GPU's are about worthless, then yeah that's a thing too.

[UltraMega]

I had a customer bring me his PC that had a 7900xtx and I tested it in furmark. On that GPU, the temps were very impressive. 62c under load. 

[bonami2]

8 hours ago, pioneerisloud said:

Would you mind expanding on that last point a little?  That's a first I'm hearing about that one, and I have a lot of AMD GPU's I don't want ded.  Unless you're just meaning the drivers that absolutely nuke performance to the point last gen GPU's are about worthless, then yeah that's a thing too.

Here we go

 

https://videocardz.com/newz/gpu-repair-service-reports-a-sudden-spike-of-broken-radeon-rx-6900-6800-gpus

Edited January 12 by bonami2

[pioneerisloud]

37 minutes ago, bonami2 said:

Here we go

 

https://videocardz.com/newz/gpu-repair-service-reports-a-sudden-spike-of-broken-radeon-rx-6900-6800-gpus

BRB, swapping drivers.

[Mr. Fox]

9 hours ago, pioneerisloud said:

Would you mind expanding on that last point a little?  That's a first I'm hearing about that one, and I have a lot of AMD GPU's I don't want ded.  Unless you're just meaning the drivers that absolutely nuke performance to the point last gen GPU's are about worthless, then yeah that's a thing too.

This is what I am referring to, in case you haven't already seen it. By dead I mean non-functional with the GPU core and some of the circuitry destroyed.

 

Edited January 12 by Mr. Fox

[bonami2]

Am still running latest like a moron. But am running fan at really high rpm.

 

Will be swapping to a May driver anyways to fix google earth bug in VR.

 

My experience with amd is they can't control fan speed good enough so that may be the issue. Slowly cooking the card till something fail.

Edited January 12 by bonami2

[pioneerisloud]

6 hours ago, bonami2 said:

Am still running latest like a moron. But am running fan at really high rpm.

 

Will be swapping to a May driver anyways to fix google earth bug in VR.

 

My experience with amd is they can't control fan speed good enough so that may be the issue. Slowly cooking the card till something fail.

I mean my "hotspot" has always hit 110*C on this card since I got it.  RMA'd it once and that didn't change a single thing.

[bonami2]

44 minutes ago, pioneerisloud said:

I mean my "hotspot" has always hit 110*C on this card since I got it.  RMA'd it once and that didn't change a single thing.

Same and mine was crashing driver when around 110c. Now fan run at 2.5k rpm and no more issue 

[Mr. Fox]

Hard to know for sure if it is heat related or something else. Probably not if heat has always been a challenge. Kind of scary though whatever the issue is. It can't be a coincidence that suddenly that many dead GPUs show up in a single repair shop with the same problem and all using the same driver. That's way too many to be a coincidence.

[UltraMega]

Sure doesn't seem like a driver issue. I don't think their evidence supports that theory either, as one would expect most gaming GPUs to be running the latest drivers, especially AMD since they don't release drivers every 2 weeks like Nvidia does. All they can really conclude is that most people using modern AMD GPUs keep their drivers up to date. 

[Sir Beregond]

On 11/01/2023 at 11:57, Mr. Fox said:

This has been a really bad couple of years for GPUs.

• Crypto-scalping
• Post-crypto scalping
• Supply/demand manipulation
• Unlaunched inferior version of the 4080
• NVIDIA 12+4 pin connectors burning
• Misrepresentations of 7900 XT/XTX performance
• Continued manipulation/suppression of 4090/7900 XTX supply
• Relaunched 4070 Ti (the inferior 4080) grossly overpriced
• Defective AMD vapor chambers
• AMD drivers killing last-gen GPUs

 

I likely forgot about something that should be on the list.

 

How about keeping 2+ year old 30-series cards at MSRP (at best) and trying to slot that in with 40-series to justify the ludicrous pricing as if to pretend these are concurrent generations and 40-series isn't for "the plebs"?

 

Confused about the AMD drivers killing RDNA2 GPUs, saw the article and they did mention all the cards having the same driver, but then show a pic of a cracked die. I'm at the office today, is it in the videos?

Edited January 13 by Sir Beregond

[Mr. Fox]

22 hours ago, UltraMega said:

Sure doesn't seem like a driver issue. I don't think their evidence supports that theory either, as one would expect most gaming GPUs to be running the latest drivers, especially AMD since they don't release drivers every 2 weeks like Nvidia does. All they can really conclude is that most people using modern AMD GPUs keep their drivers up to date. 

We don't know what we don't know right now. But, something is certainly not right. That is too many nearly simultaneous failures. The frequency seems too high to be coincidental regardless of what the underlying cause is determined to be.

4 hours ago, Sir Beregond said:

How about keeping 2+ year old 30-series cards at MSRP (at best) and trying to slot that in with 40-series to justify the ludicrous pricing as if to pretend these are concurrent generations and 40-series isn't for "the plebs"?

 

Confused about the AMD drivers killing RDNA2 GPUs, saw the article and they did mention all the cards having the same driver, but then show a pic of a cracked die. I'm at the office today, is it in the videos?

I don't think the exact cause has been determined. The only thing consistent between GPUs in an extremely strange frequency of identical failures was the driver version. That was actually brought out in the video. Nothing else could be identified (yet) and they are asking for info to help identify both cause and frequency.

 

That said, I do not rule it out as a possibility. We have seen numerous examples of NVIDIA doing things with drivers, including making small adjustments to firmware without disclosing it in release notes. For example, overclock blocking and removing access to voltage controls. This actually just happened with Maxwell GPUs. If you are running the latest drivers on a 980 Ti, you will likely find you can no longer overvolt or undervolt the GPU. I can still remember severely throttling on Kepler SLI systems I owned that was 100% triggered by nasty GeFarts drivers. It never got fixed. NVIDIA did not care. They were selling Maxwell. The only solution was to use an outdated driver if you wanted SLI performance, or buy two new Maxwell GPUs.

 

Those were deliberate acts on NVIDIA's part, but accidents and unintended consequences can also occur. The fractured die can result from extreme overheating or a freakishly high voltage spike, so it is possible (although maybe remote) that physical damage can result from a driver update, especially if it is carrying a special payload of "extras" in the package.

 

Purely speculation now, but not outside the realm of possibility... AMD might have released something in the December driver that was designed to push 7900 XTX performance to make it more competitive against 4080 and 4090 that had unforseen consequences for their last generation GPUs.

Edited January 13 by Mr. Fox

[pioneerisloud]

On 13/01/2023 at 13:36, Mr. Fox said:

..........

 

Purely speculation now, but not outside the realm of possibility... AMD might have released something in the December driver that was designed to push 7900 XTX performance to make it more competitive against 4080 and 4090 that had unforseen consequences for their last generation GPUs.

I was actually just sitting here pondering that very same scenario......  

[Mr. Fox]

On 14/01/2023 at 18:57, pioneerisloud said:

I was actually just sitting here pondering that very same scenario......  

I think the exact cause is still not verified, but the last video from KrisFix-Germany seems to indicate the possibility that most (if not all) of the affected GPUs came from the same vendor in China that may have been selling retired mining GPUs. I believe that all or most of them worked for a while in the hands of the buyer before the cores fractured. They were all abnormally clean so it may have been something with the cleaning/refurbishment process that contributed to the failures. We may never know for certain. Kris had to dig the information out of the customers that were willing to provide information to help with his investigation of the cause. Some were unresponsive.

Edited January 30 by Mr. Fox