More trouble for AMD GPUs ? Caution using the latest Radeon drivers (6800XT, 6900XT and more)

[J7SC_Orion]

...in case this hasn't been covered here yet, the two related vids are self-explanatory re. even more potential trouble for AMD's GPU division. For what it is worth, I rolled the driver on my 6900XT back - while it didn't blow up (unlike other GPUs) and my GPU is extensively water-cooled anyway, there were definite signs of trouble and unexplained weirdness with the latest Radeon driver...since I rolled it back, no more issues (at least so far).

 

 

 

[bonami2]

Interresting in the comment people talk about people possibly installing the 7900 serie driver because the website was buggued or something when it was released.

 

Am running latest at 100% load f@h without issue on my 6800xt.

[J7SC_Orion]

21 minutes ago, bonami2 said:

Interresting in the comment people talk about people possibly installing the 7900 serie driver because the website was buggued or something when it was released.

 

Am running latest at 100% load f@h without issue on my 6800xt.

 

...yeah, better safe than sorry - for now, I stick with the slightly older driver (early November '22). Krisfix.de doesn't fool around; s.th. seems to be going on but hopefully, this gets cleared up soon. It could also just be that folks who have been running custom MPT (up to 600W, ++ V for a 6900XT) may just have had their setup pushed over the edge by a minor change within drivers.

[UltraMega]

I saw this recently, but it seems like way too small of a sample size and geographically isolated. One guy in Germany is the only source for this whole thing now. Almost sounds like this guy is cracking dies when he removes the heatsink somehow.

Edited January 20, 2023 by UltraMega

[pio]

This WAS covered elsewhere on the site, but I'm pretty sure it was a regular thread and not a news thread.  @bonami2 I think it was your thread?  Either way, I don't mind its posted twice, but you had kinda asked......

 

Anyway, I too WAS running the 22.11.2's I think they were, the "cursed" drivers mentioned in the videos.  I had absolutely ZERO problems on them myself with a bone stock 6900XT.  With that said, I only was on them for a few days and hadn't even really fired up a game as I had just formatted and re-installed.  I read about the problems in bonami's thread, and well....I'm now back on 22.11.1's.  I believe J7SC said it earlier, but better safe than sorry.

 

For some odd reason, I actually WAS experiencing driver related problems on the older 22.5.6's I think they were?  The last WHQL's before these last 2 sets.  The problems I was having, also happen to be noted fixes in the newer 22.11.1's I'm running now.  Black screen problems, freezing, and those are notably "fixed" now.  I also DID in fact eventually start noticing those same issues on my 5700XT and my Vega64 systems on occasion (but not 110% sure it was drivers on that, so whatever).  I was NOT having those problems earlier this year though in like January - April ish drivers.  So, its kind of interesting that AMD broke the drivers sometime earlier this year, and it took them 6 months to issue a fix.  At least, for my particular bugs I was legitimately experiencing.  I think I had the black screen bug once when I was streaming "Lawn Mowing Simulator" in the site discord channel lol.

Edited January 20, 2023 by pioneerisloud

[J7SC_Orion]

1 hour ago, UltraMega said:

I saw this recently, but it seems like way too small of a sample size and geographically isolated. One guy in Germany is the only source for this whole thing now. Almost sounds like this guy is cracking dies when he removes the heatsink somehow.

 

wow - really ? 
'This guy' is highly skilled and has equipment that normal repair faculties can only dream about....just take a look at the time-marked video below (actually the whole vid is fascinating).

 

You wrote that "Almost sounds like this guy is cracking dies when he removes the heatsink somehow." - that is a ludicrous insult, not least as he has seen that problem in 48 out of 61 cards his facility alone has been sent - and they are well known in Europe for this kind of niche work. I know that Gigabyte EU also sometimes uses that firm. You are not seriously suggesting that he carelessly screwed up 48 cards ? 

 

Next, because his outfit is well known, they would receive a large segment of serious failures. What is more, he is asking online for more input from AMD users rather than jumping to conclusions. The only common factor he could identify so far is the specific driver, and the sudden rush of cards breaking is telling..

 

It is worth pointing out that AMD has had multiple issues as of late. It was DerBauer (and Igor's lab) who did most of the initial detective work isolating the problem with the vapor chambers (fluid) on reference 7900XT/X. AMD since then confirmed that early sleuthing by DerBauer et al...

 

Finally, I have been enjoying my custom-PCB 6900 XT for almost two years now. It has been problem-free - until I installed the latest driver. Since I uninstalled it, the problems are gone...I repeat: better safe than sorry.

 

Edited January 20, 2023 by J7SC_Orion

[UltraMega]

5 hours ago, J7SC_Orion said:

 

wow - really ? 
'This guy' is highly skilled and has equipment that normal repair faculties can only dream about....just take a look at the time-marked video below (actually the whole vid is fascinating).

 

You wrote that "Almost sounds like this guy is cracking dies when he removes the heatsink somehow." - that is a ludicrous insult, not least as he has seen that problem in 48 out of 61 cards his facility alone has been sent - and they are well known in Europe for this kind of niche work. I know that Gigabyte EU also sometimes uses that firm. You are not seriously suggesting that he carelessly screwed up 48 cards ? 

 

Next, because his outfit is well known, they would receive a large segment of serious failures. What is more, he is asking online for more input from AMD users rather than jumping to conclusions. The only common factor he could identify so far is the specific driver, and the sudden rush of cards breaking is telling..

 

It is worth pointing out that AMD has had multiple issues as of late. It was DerBauer (and Igor's lab) who did most of the initial detective work isolating the problem with the vapor chambers (fluid) on reference 7900XT/X. AMD since then confirmed that early sleuthing by DerBauer et al...

 

Finally, I have been enjoying my custom-PCB 6900 XT for almost two years now. It has been problem-free - until I installed the latest driver. Since I uninstalled it, the problems are gone...I repeat: better safe than sorry.

 

 

All I'm saying is one guy is not enough to get worried over. As the video you posted said, it's strange that only one person has reported this type of issue and no one else has come forward with a similar experience since then. 

 

I also don't see why suggesting that this guy could be doing something wrong is such an insult. Seems more likely than him being the only person in the world to have this experience... 44 times. 

[schuck6566]

In the comments on Jayz vidieo (and Jay himself also) people were wondering HOW the driver could be causing the cards to over heat and the memory to over heat nada,nada,blah,blah when those functions are in the bios....  A simple Google search showed that the Adrenalin Software from AMD that contains the drivers ALSO contains software to overclock the GPU,Memory,AND play with the voltages! Maybe this latest driver update affected the Adrenalin software so that it reflected the settings that were used when driver was written?

[J7SC_Orion]

3 hours ago, schuck6566 said:

In the comments on Jayz vidieo (and Jay himself also) people were wondering HOW the driver could be causing the cards to over heat and the memory to over heat nada,nada,blah,blah when those functions are in the bios....  A simple Google search showed that the Adrenalin Software from AMD that contains the drivers ALSO contains software to overclock the GPU,Memory,AND play with the voltages! Maybe this latest driver update affected the Adrenalin software so that it reflected the settings that were used when driver was written?

 

...^ that is a key point. 
Drivers with AMD cards are different - most of the bios details (including limits) are written into the Windows registry and accessible, unlike NVidia. Drivers also interact with that and that is why MPT works so well (very popular in the Hardwareluxx community with the guide located there). Take the first screenshot below of my 6900XT (water-cooled in a 1200mm x 63mm dual D5 setup for extra headroom). Stock on the left, but per right side, 520 W, 1.243V is no problem and theoretically I could go to 600W and 1.4V (could, but won't). This does NOT involve vBios flashing as much as my custom card has 3x8 pin and dual vBios (I flashed one for 6950XT / Ubuntu). There are a whole host of other settings w/MPT and driver combo you can change. If the latest AMD driver has a just a minor mistake (as Jay2C suggested just as an example / possibly a decimal point moved), you could end up with serious problems - especially if your setup is air-cooled.

 

 

 

I run various system for work (including ML at times) and play - and certainly try to find the sustainable (error-free) limits, per screenie below on just my current work-play builds. I have used both AMD and NVidia cards for well over a decade, and AMD and NVidia are very different in how they handle the vbios-driver relationship, especially on power and voltages. It even affects your CPU OC at the limits (ie. for HWBot) as one GPU type requires a touch more CPU voltages at the limit.

 

 

 

Edited January 20, 2023 by J7SC_Orion

[UltraMega]

Looks like the guy who started this whole thing just posted an update. TLDR, not the driver. The guy now believes these cards were mining cards that had been in storage for a while and not properly stored, then passed off as new cards. 

 

 

It was always obvious that it wasn't the driver because driver issues are not geographically isolated. 

Edited January 20, 2023 by UltraMega

[J7SC_Orion]

14 minutes ago, UltraMega said:

Looks like the guy who started this whole thing just posted an update. TLDR, not the driver. The guy now believes these cards were mining cards that had been in storage for a while and not properly stored, then passed off as new cards. 

 

 

It was always obvious that it wasn't the driver because driver issues are not geographically isolated. 

 

...that is some relief. But notice how thorough he was with his due diligence, which is why the original vid was posted - and no, he was not breaking cards when taking the cooler off as you suggested above 

 

I also reiterate that I've had problems with the latest driver, ditto for other folks per OC forums. The fact that some cards may have had a rough life as a mining card would underscore that there's no margin for error...that leads to another conclusion: Don't buy used mining cards, though at times, sellers might not disclose that...

[UltraMega]

5 minutes ago, J7SC_Orion said:

 

...that is some relief. But notice how thorough he was with his due diligence, which is why the original vid was posted - and no, he was not breaking cards when taking the cooler off as you suggested above 

 

I also reiterate that I've had problems with the latest driver, ditto for other folks per OC forums. The fact that some cards may have had a rough life as a mining card would underscore that there's no margin for error...that leads to another conclusion: Don't buy used mining cards, though at times, sellers might not disclose that...

 

I could have worded it better but the point I was getting at is that this is an isolated event to one place so it had to be something about that place, and it turns out it was if his new theory is right. I think his new theory definitely makes sense. 

 

I'm not sure I'd call it that thurough because he could have asked his customers more about where they got their card as soon as he noticed a trend and figured this out a lot sooner. Instead, he asked the internet to look for answers for him. 

[J7SC_Orion]

5 minutes ago, UltraMega said:

 

I could have worded it better but the point I was getting at is that this is an isolated event to one place so it had to be something about that place, and it turns out it was if his new theory is right. I think his new theory definitely makes sense. 

 

I'm not sure I'd call it that thurough because he could have asked his customers more about where they got their card as soon as he noticed a trend and figured this out a lot sooner. Instead, he asked the internet to look for answers for him. 

  

...how else is he going to reach the largest possible user group, even beyond a given region ? 

[UltraMega]

11 minutes ago, J7SC_Orion said:

  

...how else is he going to reach the largest possible user group, even beyond a given region ? 

He didn't need to in order to figure this out. 

[bonami2]

50 minutes ago, J7SC_Orion said:

 

...that is some relief. But notice how thorough he was with his due diligence, which is why the original vid was posted - and no, he was not breaking cards when taking the cooler off as you suggested above 

 

I also reiterate that I've had problems with the latest driver, ditto for other folks per OC forums. The fact that some cards may have had a rough life as a mining card would underscore that there's no margin for error...that leads to another conclusion: Don't buy used mining cards, though at times, sellers might not disclose that...

It just go back to what i said earlier.

 

Amd fan speed by default are ways too low and cook the Gpu.

Running at a fixed temp should be easier on the pcb and solder joint than gaming.

 

Mining is like running a car or server at full load 24/7 less heat cycle. Less failure.

 

 

 

 

 

 

[J7SC_Orion]

1 minute ago, bonami2 said:

It just go back to what i said earlier.

 

Amd fan speed by default are ways too low and cook the Gpu.

Running at a fixed temp should be easier on the pcb and solder joint than gaming.

 

Mining is like running a car or server at full load 24/7 less heat cycle. Less failure.

 

 

 

 

 

 

   

True, but as mentioned, my 6900XT sits in a 1200mm x 63 mm 2x D5 loop and temps have never been the issue.  The latest driver caused some blackscreening and inexplicable freezing for 20 sec or so (only in Win 11, not Kubuntu); rolling back the driver solved that on my particular setup...

[bonami2]

2 hours ago, J7SC_Orion said:

   

True, but as mentioned, my 6900XT sits in a 1200mm x 63 mm 2x D5 loop and temps have never been the issue.  The latest driver caused some blackscreening and inexplicable freezing for 20 sec or so (only in Win 11, not Kubuntu); rolling back the driver solved that on my particular setup...

I was only talking about the hardware failure.

 

Software wise i have no issue running vr or 4k or f@h and even all together because i forget turn off f@h. With my 6800xt.

 

Amd black screen is part of their driver hahah but i never had a black screen with my 6800xt 

[pio]

4 hours ago, J7SC_Orion said:

   

True, but as mentioned, my 6900XT sits in a 1200mm x 63 mm 2x D5 loop and temps have never been the issue.  The latest driver caused some blackscreening and inexplicable freezing for 20 sec or so (only in Win 11, not Kubuntu); rolling back the driver solved that on my particular setup...

As I mentioned too in my comment, I had those same EXACT problems of blackscreens and random freezing on older drivers.  It was especially prevalent when watching youtube videos, just on desktop use.  For me, PERSONALLY, I did not have those same issues on 22.11.2's, but I went back to 22.11.1's as a precaution.

Edited January 20, 2023 by pioneerisloud

[J7SC_Orion]

9 hours ago, bonami2 said:

I was only talking about the hardware failure.

 

Software wise i have no issue running vr or 4k or f@h and even all together because i forget turn off f@h. With my 6800xt.

 

Amd black screen is part of their driver hahah but i never had a black screen with my 6800xt 

 

...I was actually referring to hardware re. the ample w-cooling my 6900XT gets and thus is a bit more protected. As to software - no, I don't like blackscreens by AMD, NVidia, Intel or really, by anybody. I do have a can of Rustoleum flat black right here in case I ever change my mind on that .

 

After the vapor chamber debacle and other GPU issues such as what you referenced in > this thread, AMD does have to be very careful, not least as their new GPUs are a bit underwhelming re. expected performance. Be that as it may, I am glad that the OP problem seems to be not at their end, and by now they probably are in contact with Krisfix.de to get the serial numbers of the affected cards to track the sales history as much as they can (there are of course resales which can muddy the water).

 

AMD is facing the same economic and other uncertainties many other tech firms now face, all the while they are competing against Intel re. CPUs on the desktop / laptop and the server market, and against Intel and NVidia in the consumer and the important enterprise GPU market. That is a lot to chew on for anybody, but the last thing it needs are quality control issues right now. I do own several AMD GPUs and 3x 16 core CPUs of different gens (used mostly for work related stuff), and I really like the firm, but the last few months have been a bit rough in the quality and PR department - even before > this 

[pio]

36 minutes ago, J7SC_Orion said:

 

...I was actually referring to hardware re. the ample w-cooling my 6900XT gets and thus is a bit more protected. As to software - no, I don't like blackscreens by AMD, NVidia, Intel or really, by anybody. I do have a can of Rustoleum flat black right here in case I ever change my mind on that .

 

After the vapor chamber debacle and other GPU issues such as what you referenced in > this thread, AMD does have to be very careful, not least as their new GPUs are a bit underwhelming re. expected performance. Be that as it may, I am glad that the OP problem seems to be not at their end, and by now they probably are in contact with Krisfix.de to get the serial numbers of the affected cards to track the sales history as much as they can (there are of course resales which can muddy the water).

 

AMD is facing the same economic and other uncertainties many other tech firms now face, all the while they are competing against Intel re. CPUs on the desktop / laptop and the server market, and against Intel and NVidia in the consumer and the important enterprise GPU market. That is a lot to chew on for anybody, but the last thing it needs are quality control issues right now. I do own several AMD GPUs and 3x 16 core CPUs of different gens (used mostly for work related stuff), and I really like the firm, but the last few months have been a bit rough in the quality and PR department - even before > this 

They've been around just as long as Intel has been.  I'm not going to sugar coat it, if there's a problem with these cards then they should just come forward and be truthful about it.  I've defended the "hurr durr AMD driver problems durrrrr" threads in the past, but OBVIOUSLY there's problems in their driver department for sure as of lately.  And you're right, the PR hits they're taking as of lately definitely aren't helping things any at all.

 

They've killed performance purposely on old cards and completely dumped support for cards they really shouldn't have (Nvidia's done similar mind you).  These black screen / freezing problems prevalent on 3 generations (at least) of cards due to their drivers, that's another problem.  And then the rumors of dies cracking at random, POTENTIALLY due to drivers (but maybe not really?).  I wanted a 7900XTX, but I'm glad I waited now.  Kind of wanting to grab a 6950XT though in case my 6900 decides to join the others.

 

-_-

Edited January 21, 2023 by pioneerisloud