GPU related failures possibly due to improper storage conditions (moisture)

[pharma]

FYI ...

 

[Cyan]

can't stand jay2cents so I didn't watch the video -I have him blocked on my yt account so I can't receive video suggestions from him anymore, fortunately so-. That being said, the actual author of the investigation has studied the cards behaviour for more than 150 hours and he mentions the results in an updated video released today:

 

[WhiningKhan]

For those like me who can't stand watching youtube videos for a piece of information, an article. Frankly I didn't watch even the videos linked in this thread but article content sounds similar enough.

The Mystery of the Radeon RX 6000 Mass Extinction Event May Have Been Solved

Ex mining cards were probably a victim of humidity from poor warehousing

www.tomshardware.com

 

[pharma]

So cards stored in humid conditions for weeks/months is the likely reason behind the failures? He should check other IHV cards since crypto cards on the used market number in the hundreds of thousands.
I guess any eBay used electronics purchase item might fall into this category if stored in similar conditions.

According to KrisFix, these cards were likely stored for a few weeks or months since GPU-based cryptomining became uneconomical. The problem is that they seem to have been stored in an environment with inappropriate temperatures / humidity levels. The experienced electronics repairer says he has seen this exact symptom of chips cracking and popping up from the PCB after being used in the wake of this kind of inappropriate storage.

KirsFix explains that when he buys electronics that may have been stored in an unsatisfactory way, perhaps passing through widely variable climates in transit, he will open the items to let them acclimatize, fully dry, and settle at room temperature. These catastrophically damaged GPUs show that, even though they worked fine initially, the effects of humidity that had gotten deep into the products caused the silicon to crack during / after their first rigorous session under load.

In other words, these cards would have worked fine immediately after decommissioning from the crypto mines, but humid storage conditions “soaked” them, and without proper dehumidifying they were going to face problems.

 

[WhiningKhan]

So cards stored in humid conditions for weeks/months is the likely reason behind the failures? He should check other IHV cards since crypto cards on the used market number in the hundreds of thousands.
I guess any eBay used electronics purchase item might fall into this category if stored in similar conditions.

Yes, this applies to pretty much all electronics (though most devices have less punishing temperature gradients compared to GPU). It is not related to IHV or even crypto usage.
I work at company which manufactures electronics and this is an actual thing, in our cases power semiconductors have had similar fates. There is a reason for environmental specs for storage.

 

[pharma]

Yes, this applies to pretty much all electronics (though most devices have less punishing temperature gradients compared to GPU). It is not related to IHV or even crypto usage.
I work at company which manufactures electronics and this is an actual thing, in our cases power semiconductors have had similar fates. There is a reason for environmental specs for storage.

Makes one wonder what the actual storage conditions were. Most my old electronic items are stored in the garage where summer is very hot/humid and winter cold/dry. I haven't had any issues getting any to work after years of storage but that doesn't mean damage does not exist.

 

[Reznor007]

Often times the damage comes from moisture from the air that penetrates the chips, and then if it's rapidly heated, it will can cause the chip to sort of burst, which could explain the cracked dies seen here.

I work in electronics manufacturing, and if a package of sensitive parts is left open for too long, or improperly packed from a supplier, the parts must be slow baked in an oven or dry box to remove moisture before the part is put through a reflow oven for soldering. I've seen chips that didn't go through the process burst through the top of the plastic casing when heated via hot air rework, and bga components have their layers separate

 

[WhiningKhan]

Often times the damage comes from moisture from the air that penetrates the chips, and then if it's rapidly heated, it will can cause the chip to sort of burst, which could explain the cracked dies seen here.

Precisely, that is what I was referring to. Reflow soldering of course is the most extreme test for the components and up to that point their environment _must_ have been controlled. But also in a finished product, a working GPU can and does go through quite violent temperature surges, and materials deform or burst when moisture escapes. My work is with with firmware and software stuff, not directly with the electronics manufacturing, but I've been told and shown plenty of such events. And you wouldn't believe how many times I've heard wild associations of failures to updated software, when the actual cause of death is purely a hardware event. People are good at inventing non-existing patterns, as in the topic of this thread...

 

[Kaotik]

In this case it might be simpler than that - we've seen miners clean their farms with pressure washers and what not, which could lead to enough water seeping in wrong places

 

[Scott_Arm]

Typically if you want to save components you’ll bake them in an oven and then vacuum seal them. Populated PCBs can definitely fail if they’re stored in humid environments.