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


Lol it's on the rain these graphics? Storage conditions may vary from 0 to 90% humidity :) working conditions 60% or more. Beware radeon user never let sleep or shutdown your machine or you will have a cracked die
  • Like
Reactions: Why_Me


No one has the clues sir. I work with electronics and never see any kind of damage even with bad storage.

If this become the new standard " your warranty says good bye"
  • Like
Reactions: artk2219


First thing that came to mind for me was that earlier story and video where mining cards were literally being hosed down for cleanup in preparation for sale.
That would do it, not the hosing down, but then not letting them properly dry and having everything sit saturated for a while. Then all of a sudden, hey heres a massive heatload, I hope you dont start expelling that water quickly and have a nice steam explosion. Its still a bit strange though, you'd figure if they were that saturated there would be at least some surface rust.


Oct 5, 2018
First thing that came to mind for me was that earlier story and video where mining cards were literally being hosed down for cleanup in preparation for sale.

Video Allegedly Shows Crypto Miners Jet Washing Nvidia RTX GPUs



Jun 23, 2014
I think it may have been some mining bios/driver issue where the cards were mined for a long time with the fans unintentionally off. Maybe the cards had cracks, maybe the TIM dried to glue and held residual stresses between the chips and heatsinks. And the thermal cycling that normal consumers do was enough to put them over the edge.

Water on the die would not be an issue, dies can take LN2. And water on the pcb wouldn't break the die, the soggy pcb would be damaged..

Seems more like some substantial thermal abuse. Maybe an issue where the TIM turns to a glue, under a lot of heat and/or time, capable of destroying the chip, especially if you remove the cooler.

Edit: Or twist the cooler hard enough relative to the die. Most coolers kind of float. If they get glued to the die, twisting from the ends of the cooler could be pretty hard on that small, brittle die.

Edit again: Now I want to replace the TIM on my 6800 that is working fine and doesn't get too hot. But I don't want to break it.
Last edited:
  • Like
Reactions: artk2219


Apr 15, 2020
I suspect someone ran these through a PCB wash cycle without first removing the heatsinks. That could get water trapped in gaps between the heatsink and the IC. When it was turned, the water turns to steam and bad things happen. PCB washing is very common if you do not use a "No Clean" soldering process. Another other possibility has them removing the heatsink, washing the PCB, then putting the heatsink back on without first drying the PCB . Another possibility is they put them in an oven to dry and ramped the temperature up too fast causing steam to form between the heatsink and the PCB. The heatsink, if attached would accept the heat from the oven very quickly.


I must wonder though how he knew they were mined. Did he ask the owners where they got the card from?
40+ of the GPUs were still under warranty, so he called the customers to ask them why the heck did they send the GPUs for repair. 25 of them responded and told him they have no receipt for RMA because they bought the GPUs used online in late november through early december.

When 25/25 of people with the same issue have bought their used GPUs at about the same time, there is a pretty good chance they all came from the same batch of improperly stored GPUs where moisture was left on long enough to seep deep enough into the GPU die and make it pop under load.

If you buy a used GPU and it looks factory-clean, assume it got pressure-washed and bake it at 70-80C for 24h before use.


Jun 16, 2007
This is one of the many risks with cards that were used in mining farms. They are abused heavily, as the goal of them is to bring the TGP value as close to the TBP value as possible while trying to maximize hash rat without delving well into the realm of diminishing returns.
Since GPU mining has a time factor to it (mining difficulty increases over time as the global mining hash rates increase, this creates a GPU compute arms race between miners, thus they end up running the cards very close to their thermal limits (fan curves focused on reducing TBP), and occasionally doing things like strapping an extra heatsink on certain parts of the board to provide a little extra cooling to a few DRAM chips that may be overheating. The goal iis to maximize profits over the entire planned length of time it will be spent mining, taking into account increases in mining difficulty, and all other factors are distant afterthoughts.

The end result is cards where the GPU and VRAM spend months of years very close to their thermal limits.

They will then behave like the NFT bros when it comes to sycophantic downplaying of the abuse of the cards, and try to make it like they downclocked the card by 80% and kept mining farm on Pluto to ensure that the cards would be kept cool. Though this is because they have a lot of money riding on a card that they minded on for years and then tricking people to spend close to MSRP for their used and abused cards.
This is similar to used car dealerships that attempt to resell a fleet car or "Program" car, which are card used in rental car fleets, where they are abused and rarely serviced before getting a bunch of maintenance done at the last minute before being sold. The sucker who ends up with the cae finds out a few months later wen they go in for an oil change, that the car burned 2+ quarts of oil and has lost most of its compression, before doing a surprised pikachu face when the mechanic wants like $10K to fix the engine.
  • Like
Reactions: artk2219