Latest driver killing 68xx/69xx ?!

Probably.

 

100% miner cards. That'd be the only reason they sent them to a aftermarket repair shop instead of sending them in under warranty of which they were still under warranty last month

 

Could be it, all in the same area, maybe all from the same mining farm.

 

cracking wouldn't be caused by mining, its a heat flex result not a constant heat result.

 

Whats the possible causes
Too much voltage?
External pressure on the die?
Structural failure due to repeated extreme heat cycles?

 

Yes, but rapid cycles, like being run in a sub-zero ambient temp with an almost frozen water cooler.

somethings sus about all this, Kris is just bringing to light the information he has based on all the cards being brought to him.

 

For sure, ive been into this more seriously since maybe 2000 and i dont think ive ever known a group of cards to end up like this.

 

Actually, I've been curious about this, but what's stopping them from sending the cards in for RMA? Is it due to the VBIOS version being a modded BIOS for mining?

 

The miners may have modded the cards for enhanced performance with component mods / custom firmware which they cannot remove.

 

If it was true, then the driver must be so bad that it makes the card to operate way above specs. This is not the case. So it is false rumor.

Such problems are hardware related, meaning that the graphics card manufacturer did something wrong in the design of the card.

So, use the latest 22.11.2 WHQL without worries, as I do.

 

It makes no sense honestly if it suggest 2479 mhz and i apply 2479 i get 2555 mhz while getting 2468 mhz when i set 2478 mhz something clearly not working the way it should.
heck auto oc sets target of 2550 mhz even more odd so card is overclocking on stock.
Btw software conflicts can kill hardware as well on an Asus X570-E with older vision of openrgb it can bug the chipset fan to a point it stops spinning, on newer visions it only pauses for a second and resumes not sure if it always resumes, im not touching that software ever again.

on 22.5.1 stress test only boost 2200 to 2300 mhz and go's full wattage while on newer drivers it boost to 2524 2530 constantly spiking up and down in load while using 50 watts less i would't be so sure.

Anyway if indeed cards getting killed there will be more dead cards ariving soon probably at other repair shops as well probably as well, i don't think AMD can hide this cos people already quite fearful of upgrading drivers these days so people might be unknowingly stay on drivers that are broken.

 

the affected cards are completely stock.

 

It makes sense. AMD cards have multiple boost tiers based on temps and voltage, it's strange how it shows a "limit" in the tables but AMD limits are more of a suggestive limit than a hard limit

 

I think its been this way for a while, the software seems to be over ridden by the cards sometimes

 

The clocks have step tiers. When the clock is set to a number, it will fall into a step tier of performance. So 2478 and 2479 is a dividing line between steps. Working as expected. Other makers do the same thing.

 

Probably some large mine has been liquidated and the owner is trying to minimize losses, and NVidia's marketing is trying to bake a new cake.

 

The story is certainly gaining traction in the press.

 

And there's been no reports of any others since, from anywhere. So it's looking more and more like a mining farm looking for repair to resell

 

If yet to hear anything come directly from AMD, they communicate via YouTubers instead of communicating them self.
Another video now not from Krisfix tho.

 

This only applies to reference cards right? If so this is why I never buy reference cards.

My XFX Radeon RX 6700 non-XT is my baby.