Earlier this month, there was a widely reported story regarding a large batch of AMD Radeon RX 6000-series graphics cards (all Navi 21 models) that had mysteriously but catastrophically died. These were some of the best graphics cards, up until the latest generation parts started launching. German electronics repair shop KrisFix.de (opens in new tab) received 61 broken or malfunctioning RX 6900 / 6800 family graphics cards and found 48 of them suffered from physically cracked GPU silicon. The mystery regarding these ruined GPUs may now have been solved, with the likely culprits being the terrible twosome of crypto mining and high humidity storage.
Naturally, readers of this AMD GPU silicon cracking news were concerned about their own lovingly cared for RDNA 2 graphics cards, as evidenced by this Reddit thread. A lot of chatter focused on the Radeon driver version in use by the deceased GPUs, pondering over whether AMD had made some kind of coding error that would cause this silicon destruction. Tom’s Hardware didn’t report on the original video, as there were many unknowns and we lacked clear details. We figured it was probably a batch of abused crypto mining cards that had been sold off to punters in the vicinity of this particular repair shop. It seems we were mostly correct.
The original tale is derived from a YouTube video from KrisFix, the eponymous channel of the German board-level repairer. In an update video today, KrisFix comes to a firm conclusion (starting 18 minutes in, video embedded above) about what exactly happened to the 61 cards that he received from various customers (note that it’s not clear if these were all from one customer, multiple customers, or potentially 61 different individuals).
He reckons most if not all of the very new looking graphics cards that he received in his shop in recent weeks were bought by individual customers from a batch released by an ex-cryptominer. Perhaps there was a local eBay or Facebook Marketplace ad selling them cheap.
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.
Image 1 of 2
Readers need to be wary of the used GPUs market, but the post-crypto world has been both a source of great bargains and ticking time bombs with regard to product durability. Miners will go to extraordinary lengths to clean up and sell on their old GPUs, but thankfully we haven’t heard too many tales like this one from KrisFix.