Hey AMD, I am posting this NOT because I believe it, I don't believe a driver cand kill a card, unless it somehow stop the fan completely, but here's something making the round in the various forums I go to:
I thought I'd better post it here for awareness, There've been talk that these cards may still be under warranty (some parts of Europe might have 3 years warranty on cards), which begs the question, IF some of these cards may still be under warranty, why send them to some random shop for repair? Seems to me RMA is a viable option in such cases....
Edit - Other than that one time when nVidia released a driver that killed some of their cards (I think it caused the fans of some cards to stop spinning). Very seldom does a program on a driver level kills a GPU. but you can brick a card with a bad VBIOS flash. Still, I do believe that these cards were perhaps previously used as mining cards that were resold. And those who'd bought them are the unfortunate ones facing this issue. Last driver I'd had installed for my 6900 XT was the Adrenalin 22.10.2, I'm pretty confident I'd have no problem with this driver dated 08/12/22 (I think it the Adrenalin 22.11.2)
Solved! Go to Solution.
I couldn't edit my original post, so I guess the update will have to be here. Kris, the repair guy in my first post, had come back to verify that the driver (presumed to be Adrenalin 22.11.2) is NOT at fault here. This turned out as expected, that the driver wasn't the cause of the issue, and that these cards were probably used by a miner or miners, sold to these unsuspecting buyers.
An update, suspicion is on Adrenalin 22.11.2, but I find it hard to believe that anything on a driver level can break a card, unless it's a massive bug...still, IF this were true, why is there not any complaints from owners of thousands upon thousands of cards that would be affected by this issue?
Game called New World killed some Nvidia GPU's not so long ago, so I suppose it is theoretically possible
If this is a real problem, my guess would be that GPUs which follow reference design will be fine, but when there is a modified version with modified VRM, modified VBIOS and whatnot, there is a possibility that something does not go as planned. Basically it would not be AMDs fault (unless problem is with reference design), but whoever made the GPU.
Most likely, if you are worried about your GPU, make sure it is not overclocked, rather undervolt it and set some limits to max clockspeed (maybe see amd.com, what is the max gameclock for you CPU and limit it to that) and I doubt you have any risk.
EDIT: This is a good link about the Nvidia/New World -case: https://www.youtube.com/watch?v=kxoXbfzP5BU
Here's a vid from a guy I've seen a few times, he seems to be reliable, and his explanation as to why some of the RX 6800/6900 cards have died:
Basically, he thinks (and I agree to some extend) that it was users' fault, that Adrenalin 22.12.1 which was meant for RX 7000 series cards, was able to be installed in rigs with RX 6000 series cards (AMD's fault). Adrenalin 22.12.2 fixed this oversight in that it won't install in any rig that doesn't have the RX 7900 XT/XTX...
I know, when I saw the 22.12.1 driver, I knew it was for RX 7900 series cards only, but sometimes, some users just don't pay attention...
I remember the day I got my 7900 XTX reference card: I have been using a 6900 XT reference since almost two years and wanted to install the 22.12.1 drivers in order to have up to date drivers right away once I install my new 7900 card. However I wasn't able to install these drivers with the 6900 XT card: The AMD software prohibited an installation stating my hardware isn't compatible with those drivers. I wasn't able to install them - once I put my new card into the system, I was able to install those drivers right away. So personally I don't know how some people were able to use the 22.12.1 version for their RDNA2 cards.
Hmm, seems like they fixed this bug even with 22.12.1 with pushing fixed version on download server.
Because i know it was possible to do, and people with RDNA2 cards checked on drivers to see differences (there weren't many, tbh...)
i just read a thread on this forum about a 6900xt where the person installed "december" drivers ... and having temperatures issue , only "december" driver we know is 7900 drivers..
edit : what i say is wrong; 22.11.2 where release december 8th.
I'd always take complaints about higher temps with a grain of salt. Largely because the drivers always revert to default settings when you install new ones. So if you've adjusted your fan curves they'll go back to default. I see way higher temps every time I update for that reason. Then go "Doh! Reapply my saved settings." At which points the fans spin up sooner and harder and my card runs cool again.
As for the drivers, I downloaded the 22.12.1 version from AMD and ran it for a short time, then updated to the 22.12.2 for my 6900XT and noticed no issues with either. In fact the 22.12.2 has been very stable. No grey screens, no crashes, no weirdness at all.
I guess its possible for a driver to destroy a card as software can destroy a card too.
Overclocking is controlled by software now, voltages, power limits, etc. a bug in drivers could destroy a video card.
The New World issue is opening my eyes to software issues. Makes me want to monitor everything in my card when opening up any new game.
Voltages, Frequencies and Power limits are capped by VBIOS. And reflashing it without having signed one is impossible currently for RDNA 2
Technically you can avoid power limit cap, by changing base power limit (because you cannot change offset in % values)... You also technically can avoid voltage cap, by enabling feature that is disabled by default and MANUALLY dialing in min/max voltage values. Because if you won't dial them first, GPU won't even boot into system. But there is currently no way to avoid frequency caps.
Driver have no way to change SPPT. It can only read values from it to refer in Wattmann. And those values driver can read, won't be applied completely unless you actually use them in Wattmann. (those values which aren't being used in Wattmann apply on boot)
By default SPPT is empty and values are being loaded from GPU VBIOS. To actually load with SPPT on you need to clone .rom, make table of values from it and create registry folder with dynamic values based on .rom... Only then you can change them.
So calling driver for possibly doing this is BS. Driver have no write access to SPPT registry. So if people broke something with SPPT, it is clearly done by them.
It is unlikely, unless they did something on their own and didn't admit. I don't say it is impossible, though, but imo, cracked core chip is a bit too much to blame on driver only. Too severe of physical persisting impact. Usually electrical damage will not be seen on core or VRAM chip (unless it cracked first and then shorted). And thermal expansion issue... Don't remember seeing it happen on GPU's in quite a while. And for such massive case at once? Sounds unlikely.
Btw there is another beta 22.11.2 driver that is used only for testing purposes and includes VULKAN video engine. So, there is that. But it is not quite easily accessable (you must know where to look for it)
Tbh, anything could happen. But i would prefer results of deep investigation.
If seen openrgb pause my chipset fan to a point it went from 60c to 75c in seconds the moment i fired it up its been a known issue as well, people underestimate what can happen, if seen gtx 480 even get killed by drivers and a driver being pulled by Nvidia within an hour of problem being reported.
But i guess for AMD it takes months to read bug reports rather then hours, i guess they think its fine for cards to suddenly die due bad drivers, especially right now when the warranty of most cards is expired anyway.
https://youtu.be/GuH42H7AdGY remember i recorded it still when i experienced bug and wanted to see if reproduceable.
You still wanna trust AMD drivers tho after so many driver issues they had and think they aren't capable of destroying cards ?
The fact they remain silent after these reports only makes it more suspicious.
AMD periodically resets the GPU to fix the DirectX12 GPU 100% issue.
Some cards go into a deep idle state and power off.
This causes the VRM and GPU to reboot and repeat spikes.
When multiple GPUs such as mining repeatedly turn ON/OFF at the same time, a very large spike occurs on the +12V line, which may damage the GPUs.
The behavior differs depending on the firmware (BIOS) of the GPU card and the VRM controller used.
Cards without a monitor connected are more likely to enter deep idle, so be careful.
*Not all cards have problems.
I loaded the INF in Device Manager.
As soon as I apply the old INF, the anomalies stop happening.
The monitoring software never freezes or reboots randomly.
Radeon Control Panel no longer stays busy for long periods of time.
It also improved the PC becoming unstable.
No spikes occur. No problem at all.
# test system
OS: Windows 11
M/B: MSI B550 Tomahawk
GPU1: Geforce GT635
GPU2: Sapphire Radeon RX570 Mining QUAD (Minitor Not Connected)
AMD periodically resets the GPU to fix the DirectX12 GPU 100% issue.
Even if it is like that (which i am not sure because i see exactly 0 spikes in voltage or power draw on your examples) this issue ONLY affected RX 500 series.
But cards that failed are RX 6000 series.
More than that. I may surprise you, but GPU never draws consistent amount power. Every frame is a transient. And if load is high, it will just take more. With low FPS and high power draw games it can cause crashes for both transient response voltage instability (which is rare if you hadn't messed up with voltage curves), or it can cause PSU shutoff, which happened quite a bit with Nvidia cards.
Frequent reboots can damage the fan control IC and the fan due to inrush current, so I reverted to an older version of the driver.
(A fan failure can damage the GPU)
I had a similar problem with my ASRock RX6600XT.
Occurs when connected to a second PCIE.
It doesn't happen when I plug in a dummy HDMI connector (EDID).
Occurs when removed.
PC6 state (Deep idle) is broken if no monitor is connected.
Something seems too sensitive.
The same is true for MSI and Sapphire.
However, the ON/OFF behavior is different.
MSI occurs at longer intervals.
It doesn't happen with older drivers.
Fan failures shouldn't damage modern GPUs since they should thermal throttle just like modern CPUs. It will mean your FPS tanks and the heat will be maxed until you sort it out though.
The question maybe is, what controls thermal throttling? Driver or vbios? If it's the driver then it might be possible to brick the card that way as the driver could ignore thermal limits and burn the card out. But that seems like an illogical design, it'd make far more sense to control thermal limits in vbios.
even if the fan where stopped the card would still throttle etc ... i think once again these are user fault , as there is only a limited amount of card that "died"... the explaination about wrong driver able to be installed coming from "ancient gameplays" seems a lot plausible, 22.11.2 are ok ... they surely tried to install 22.12.1 (7900 drivers)...
Im sure you right, and i am sure you wanna test this out for science right just unplug your fans on your cooler and watch card protect it self right ? anyway don't do this obviously.
"unplugging" that's what you do when you install waterblock for instance, no need for fans to be plugged in..., right ?...depending on temps ... nothing magic here
I'd asked around in the forums I go to, for owners of RX 6800/6900 series cards to report on similar issue. The forums I go to are some of the larger ones with a good number of users, so far, only one had reported a positive experience with the Adrenalin 22.11.2 driver, and NOBODY else has reported anything negative.
This leads me to conclude that this may be an 'isolated' case particularly to that region. Possibly modded mining cards sold to buyers. That may explain why these card owners are seeking his help because AMD won't repair or allow for RMA of modded cards.
Not impossible like others already mentioned, i would take this potential serious AMD is't exactly known to fix driver issues fast either, their have been driver issues for months now without AMD acknowledging some of these issues, while having fixed some of these issues.
Its a bit problematic since AMD decided to stay silent, i'm already to afraid to install new driver because of driver issues, luckily i am fine running linux.
driver log "known issues" list is not what i call staying silent , before considering a "bug" you must be sure it is , ... and when so much fake rumors lately is not helping ... maybe not answer to every troll in not a bad thing
and about correcting each individual problem of every user , you can see on forums, most people accusing driver or gpu often ends up by updating bios , psu , changing cables, etc most are not related to driver or gpu but only about computer knowledge basics (computer is no game console) .. everything is relative ... and when i see all trolls and rumors spread lately that are just clickbait .. got to read carefully and ask good questions because taking things "as is" from some youtubers and reviewer is problematic (never asked yourself why some have walls of RTX boxes behind them ? just decorative ? yeah sure .. lol )... you should be careful to what is plausible when you see so much superlatives (that should ring a bell) in video title and review titles youtubers wants to make views with sometimes nothing relevant... i can give you examples if you don't see what i mean ......
and in other cases , sadly i think some people never think of using bug report tool in adrenalin imo
Sometimes, you'd read about a poster (happens often enough) complaining about bad AMD drivers, when the fault lies between the chair and the monitor. Sometimes, it's failure to update mobo chipset driver for better GPU support. You'd find such posters screaming bloody murder that AMD drivers are at fault. When the solution is finally offered. via a simple chipset driver update, they'd still act like it was AMD's fault for not releasing a driver with better compatibility. Sometimes, the chipset driver and/or firmware (BIOS) needs updating to support newer features and GPUs.
Sometimes, they'd buy mining cards (or sketchy cards sold at eBay and such, with no real info about its usage and history), then complain about cards' inability to run games well, when the solution is to flash the VBIOS to stock ones. Many a times, I've seen peeps complaining about driver issues, yes, sometimes there may be legit issues, but I think a good number of them are just down to users installing one driver over another, or simply uninstalling old driver and installing new ones.
I've used both AMD and nVidia cards, neither are perfect, as issues would arise at one time or another, but a vast majority of the time, driver install and performance have been pretty good....for both camps. I usually run DDU before a new driver install, just to eliminate any possible issue that might crop up due to bad driver install.
Oft times, shills would blow up the issue with one camp, or another, though I'd say that the net is full of peeps who've not tried any AMD cards, but based their opinions on what they'd read on the net. They'd be simply parroting issues that they might have read about and have had no experience at all. When these peeps or shills defend, say, nVidia, they'd almost always fall back on ".....but, AMD drivers sux!" argument .
As I'd stated, I use cards from both camps (have yet to try Intel GPU) and can honestly say both are good, sometimes an issue or two may crop up, just report the issue and hope for a fix with a newer driver. This happened when I was using Chrome to play a vid online, the screen would go blank/black, but a later driver resolved that issue (for AMD).
The only time I'd heard of a driver killing cards was that time when nVidia released a driver that somehow made the fans on their cards stop even under load, that did kill a few cards IIRC. As for the issue right now, I don't think it's down to the driver, as there'd be certainly many more who'd be online complaining about their RX 6800/6900 series cards being killed after the driver update.
The number of times I've helped friends fix problems which ultimately break down to "I chose the cheapest components I could find, why doesn't it work properly?" Or worse, buying one super high end component (usually the video card) but combining it with a cheap and nasty case+psu combo that cost $45 total. Then wondering why it'll randomly crash whenever it's under load.
PC ecosystem is both amazing and frustrating for the same reason - you can buy thousands of possible combinations of models and brands for components. Sadly that results in a game of "does part X work well in combination with part Y". And that's just made worse with the glut of second hand parts on the market that have been hammered in mining rigs for a year and are already on the verge of exploding into flames.
I purchased a second hand r9 390 and it worked well for 5 years of constant gaming. It had driver issues where it wouldnt reach max core speed it would always be in ulps mode with 2 monitors. Just had to disable ulps and put the clock at a static mhz. Eventually this bug was fixed.
It was for my nephews computer and it had a cheap 50 dollar case with atleast 3 fans though.
At the end of its life it wouldnt go into 3d mode without a black screen. Driver support for that card was already discontinued and we just had to call it quits. So he got another computer. This time all high end stuff. But he transferred the ram and harddrives.
I've never had a horrible experience with used cards. Just gotta know stuff. Every month i was over at my nephews checking to see if amd fixed the ulps issue. lol
JayTwoCents did a video now and i like how he also mentions driver issues, he really recommends rolling back drivers atleast until more is known about what caused this ignoring the click bait he makes some valid points
Maybe better be safe then sorry these cards cost 1000 euros more or less.
Again Nvidia has killed gtx 480 in the past almost did it again with gtx 1080 pulled drivers within an hour.
Software can even cause conflicts and issues to a point a chipset fan that has no fan control to stop spinning happened to me with open rgb like previously mentioned in a post.
Please do take it serious there could be changes to voltage that are not visible to normal users like voltage rails like AMD having made a typo and now its damaging or killing cards perhaps ?
Could even be some cards that have issues now with drivers already being perfectly binned and now drivers are damaging them or killing them ?
The original video poster, Kris, has come back with his theory as to what had happened (in the link campcreekdude had posted), we can breathe a sigh of relief as it has nothing to do with the driver. My assertion that the cards were from miners, or at the very least, shady source(s) ....Kris posited that storage and humidity has a part in this as well.
That the driver wasn't the cause of the issue was a certainty in my mind, I'd stated in another forum that had it been so, there'd be tons of users complaining that their cards had died after the driver update. As I'd mentioned, I'd asked around, none of those in the forums I go to have reported any issue with Adrenalin 22.11.2
Original driver as downloaded with adrenaline software on 30th Nov? Same here. Massive coil whine (hearable through headset) to complete computer crash, incl. unlimited fps in 3d apps despite limiters set in adrenaline software. Complete system redo a week later fixed. See
Appearntly drivers aren't killing gpu's, but try telling that to users who experience blackscreens and other driver issues since 22.5.2
I am glad i am watercooling my 6900 XT should help extend the life of my chip hopefully long enough to skip one more generation
That's the thing, both nVidia and AMD have had issues crop up from time to time. I can tell you that from the time I'd gotten my Sapphire Nitro+ RX 6900 XT, back at beginning of January, 2021, till now, I've had only minor performance issues,, and that Chrome vid playback resulting in black screen (which was resolved after a driver update or two).
That's why when I read about some having issues, I can't identify with them as I've had a great experience with a similar card (that is, 6900 XT, 6800 XT cards). My RX 6900 XT has been more or less a joy to game with, only game it'd let me down was Metro Exodus PC Enhanced Edition, RT performance isn't quite there yet.
Although I hope to resolve the RT performance issue in ME PC Enhanced with the RX 7900 XTX, with its improved RT performance. I don't particularly care about RT, but I do like ME PC Enhanced, and it's the only game I care about that has RT.