AMD's Drivers Are Killing GPUs?

0:00 - Intro

Presenter: Brett Sticklemonster
Videographer: Brett Sticklemonster
Editor: Catlin Stevenson
Thumbnail Designer: Reece Hill

[dano]

idk, I kind of doubt a driver could cause that unless its pushing massive voltage or current where it shouldnt be.

[AshtonCoolman]

I need a lot more convincing that this problem is caused by drivers than some speculation. We don't know anything about the environment these GPUs were used in. Were they mining cards? Were they overclocked? If it is actually caused by a driver alone with no other variable, I'd expect thousands and thousands of failures which would make the community worldwide A LOT more vocal than this. It would be absolutely huge by now. Disclaimer: I have no skin in the game. I own an RTX 4090.

[ZSchrink]

The VALL-E stuff is very surprising; I'm surprised it doesn't need 60 seconds or something though. That feel like such a huge leap!

[flamingscar]

Let's not panic instantly

It's possible that it's not related to driver issues across all of AMDs 6800-6900xt cards and just 1 specific bad batch of these cards

The fact it's only 6800-6900 cards lends credence to this, showing that possibly the first batch of rx6000 series have the issue, since the first cards to come out were the 6800, 6800xt, and 6900xt

And looking at the rx 7000 series shows that AMD is prone to having issues at launch

[mkennedy]

It's definitely a sign of major overheating to crack the die. Driver, user, or manufacturing error.

[UltramarinePrimaris]

Where did these GPUs come from ? What were they used for ? Why has only one youtuber(KrisFix) reported this ? Valid questions to ask before we sensationalize the issue.

[demrasnawla]

It seems that what happened was that people were able to install the 22.12.1 drivers on their previous-gen GPUs, and those drivers are exclusively meant for the 7900 series, so if you use those drivers with a different GPU it'll give it too much voltage etc. and fry it. AMD's at fault for allowing users to mistakenly install those drivers on non-7900 GPUs. The issue was fixed with version 22.12.2, which prevents users from making that error, but it was possible on 22.12.1 for a couple of weeks.

[sjbender]

AMD driver issue might be because of the noise suppression. I noticed that mine was pegged at 97% utilization at idle and found that when I was undervolting I could get the power draw down to 15w but the utilization was unchanged. After turning off the fans, LEDs, and then finally the sound suppression turned off allowed the card to rest. Power draw at 9w and utilization at 0%. Temps seemed to be in check at 65c(core) and 76c(junction) before turning off noise suppression and now at 30c at idle.

[dredgewalker]

A careful investigation is necessary before making assumptions. Even if the cards all have the same driver there are probably other similarities with the other cards. I'm all for shitting on amd if they really released a driver that destroys video cards but we really need more info and tests to replicate the issue and prove it was the drivers. Though this is also the first time I've heard of it happening. For that part to crack it needs to heat up a lot and then cool down fast.

[Savitarax]

The only thing I can possibly think of is the chip is getting extremely hot. Like the thermal throttle protection failing and then the voltage goes insane and sends 12volts to the chip at the same time.
Cause literally 12 volts wouldn’t do this.
But a 150c chip or 200c chip that’s basically melted and like glass.

[igormarcos]

How can they screw so bad to the point of a software physically cracking a chip?

[andyhill]

If you were to do the Linus video now, I would still be interested in how the new tech functions, as well as how hard it may be to actually do any of this. Show us how to Chat GPT...please?

[justuseodysee]

Considering shorted rails, chip cracking probably isn't the cause but the result of the extreme overheat in the shorted points

[Krykis]

My Gygabyte RX 6750 XT keeps crashing Adrenaline on Youtube videos several times daily. It did it in Windows 10 and a fresh install of windows 11. It has also crashed with Guild Wars 2 a few times, in everquest I only get 10 FPS when in the guild hall (lots of players there) (and when I zone it takes a several seconds to get all the buildings, landscape and players to be drawn). It also crashed a few times with The Witcher 3 and Death Stranding. City Skylines takes several seconds to draw when I first zoom into the city section. Card runs very cool, it's in a HAF XB Evo case and never hit above 47c.

[ПостороннимВ-йю]

When a video chip is reballed, it is heated with a heat gun to something like 200C. And it works absolutely fine after being exposed to that temperature. So in order for the chip to crack and fuse with the cold plate, the temperature should probably be 300C or more. I'm not sure that it's technically possible, considering the hardware level overheating protection.

[racechronic]

maybe they are cards used for mining? and other bought them and used it, which may have caused the issue due to changes made in voltages or any other things during mining state.

[albertko]

When I heard ChrisFix, I thought for a second I was wondering when a car YouTuber transitioned to tech... It's KRISFIX-Germany 😳

OHHHH 🤦‍♂️

[DigitalConceptz]

Wow!!!! That’s crazy man I was thinking about purchasing a 6000 XT or later hopefully I am safe I have 5700XT lol

[ktpt]

Well that's one way of pushing your customer to buy the 7000 series.

[OryxAU]

For those unaware, the likely cause in the opinion of the person who found these issues is that these were mining cards that were stored improperly after they were no longer in use. It's not an AMD problem.