PC Graphics

[Krystallize]

Hey guys,

so I bought my RX 6900 XT Reference Model around 5 months ago and since then I'm having frequent crashes,

and I simply can't pinpoint the issue.

There seems to be some kind of pattern to the crashes but sadly I can't reproduce them.

-During gaming, the drivers crash either very shorty after getting into the game or in some graphic intense situations (explosions)

-My drivers often seem to crash while streaming on discord and recording through OBS can even lead to a BSOD including graphicbugs on the bottom of the screen.

-Overall it seems like the drivers seem to crash more often if I turn off the 0RPM feature, which by the way let's my card get up to 60°c before even waking the fans (although the fancurve says that it should start at around 50°c)

Throughout those 5 months I did around 5 clean windows installations, used DDU more than I can count and tried different AMD Driver versions. Bios has been updated.

Memtest checked my ram, all of my storage works, my CPU got stresstested for 2 hours as well as my GPU and no issues at all.

Temps are all fine and my PSU works as intended.

My Setup:
CPU: Ryzen 7 5800X @ Stock (Dark Rock Pro 4 Cooler)

RAM: 32 Crucial Ballistix @ 3600Mhz

GPU: RX 6900 XT reference @ Stock

PSU: Be Quiet Straight Power 11 850W Gold

MB: ASUS ROG STRIX B550-E

Windows: Version 10.0.19043 Build 19043

Bios: Version 2423

SAM: ON

I genuinely don't know how to proceed, so RMAing would be my last option, but maybe you guys still have an idea :3  Oh and in case you know how the RMA Process works, is there a way to keep the videocard until you receive the new one ? I'm dependant on my card for work which is also the reason why I didn't RMA it yet (well, and the fact that I still hoped I can get it fixed :x)

If you need any more info, eventlogs for example, just tell me :3

[FederaluL · Accepted solution]

are you using MSIAfterburner or MorepowerTool? or custom GPU values in Andrenalin software?

I had the same thing with my 6900xt ref and finally solved it after I replaced the PSU 2 times. let me explain..

According to Be Quiet the Be Quiet Straight Power 11 850W Gold 12V rails outputs are:

+12V1 (A) 21
+12V2 (A) 21
+12V3 (A) 26
+12V4 (A) 26

Your GPU under load can draw 300W - 400W. the crash happens when the GPU needs more power but the PSU cannot give it. and is simple math. 

Ex: 350w at 12v = 29.17A -> but the Max your 12v rails can give is 26A. 

Do you have your GPU power cables splitting from 1 power rail? if yes, try running 2 different cables from different rails, or just get a more powerful power supply which can get you at least 40A on 1 rail.

Check your windows Event logger after a crash happens, see if there's any errors regarding Kernel Power.

I had the same problem with Straight Power 11 850W and 1000W. Solved it with a Dark Power 12 1000W.

[Krystallize]

Thanks for your suggestion, I'm pretty bad if it comes to electricity but I get what you mean.

I still have a few questions:

1. idk anything about power management of the GPU, is it posssible for it to even crash during idle (watching yt videos, DC call) due to a lack of power ? Because I had that issue during calls and on older games like dying light.

2. The power cables for my GPU are splitted by default, is this specific to those types of PSUs or is there something I'm missing ?

3. I checked my eventlogger for the last crash and found: "Beim Starten der Sitzung "OWEtKernelLogger" ist der folgende Fehler aufgetreten: 0xC0000022." which roughly translates to "while starting the session "OWEtKernelLogger" an error has occured: 0xC0000022." is this what you meant or should it say something about power ?

4. How am I able to determine what PSU should work for the card ? I'm willing to buy a new one but at least I wanna be sure to eliminate that issue at least. (I checked the manual of mine but I have no idea what to look for exactly)

[amdman]

"2. The power cables for my GPU are splitted by default, is this specific to those types of PSUs or is there something I'm missing ?"

You will want to look very carefully at the documentation for your power supply, as FederaluL was mentioning, it has now become very important which PCIe cables you plug into the video card.  See this thread for more information, look at the images they posted from their documentation, this is the info you will need for your Be Quiet Straight Power 11 850W Gold PSU:

https://community.amd.com/t5/graphics/sapphire-pulse-rx-6600-xt-black-screen-driver-instalation/td-p...

[ssaradpon]

I agree with this, as this has happened with me. To the OP try switching your old PSU with a new or known working PSU.

[ThreeDee]

Are you running separate power leads to each power input on your GPU?

Are you running the latest AM4 chipset drivers from AMD.com ?

Is your RAM actually running at 3600? (Task Manager>Performance>Memory) 

Are you running 2 x 16gb sticks or 4 x 8gb sticks?

Is your fclk set to 1800?

Do you get crashes if memory is set to default (for testing purposes) .. so it would be at 2400'ish with your ASUS board.

Have you tried a different Power Supply?

How is your airflow in your case?

In my wife's setup . .I was having driver timeout issues (crashes) with her RX 550 2GB. (5600x, ASRock B550m Phantom Gaming 4, 2 x 16gb 3200 )

Changing out the power supply corrected that issue. Thermaltake RGB 500wt 80 White (it was a highly rated "white" PSU, lol) to a Segotep (popular in China) 650wtt 80+ Gold.. 

Your be quiet! PSU is an ok power supply .. but maybe it's just not holding up consistently with your 6900xt .. especially if you are trying to power your GPU with only 1 power lead ... maybe

stuff to try/do/think about if you want .. hopefully you get it figured out and it's something simple that doesn't involve RMA'ing your GPU

[Krystallize]

Yes (unless there is something I'm missing, futher explained in the other answer :D)

Yes

Yes

2x16gb

Will check, probably in the bios ?

Yes, I also tried with and without SAM, didn't change anything

Nope, I always assumed it should be more than enough but I didn't know too much about it

(Will try out a new one if I know what to get)

Airflow should be perfect, I didn't change any fancurves for the casefans, but I'm using a BQ DX500 and checked for the best fan setup for the best balance between in and outtake.

One additional thing I just remembered, since Cyberpunk 2077 was crashing pretty frequently especially at the beginning of the game, I tried booting up my pc, immidiately started CP2077 and spammed enter and space to get into the game asap and it immidiately crashed with a BSOD and graphical bugs before even loading me up into the game. That was the firs time I tried to reproduce the issue. sadly it didn't happen again after trying it 2 more times

[TimRed]

I have an older card and have a similar issue, so I'm curious, how does it crash?

Does the system BSOD, restart, or does the screen go black with sound?

[Krystallize]

Most of the times it's just a driver crash resulting in resetting wattman settings but sometimes I get a BSOD where an eigthth of the lower part looks like my normal screen getting pressed together.

[ThunderBeaver]

Personally I've never used the fan curve setup only manual.

When not gaming I run my GPU fans at 60 to 75% for reduced noise. 

When gaming I run at 90 to 100% because heat kills efficiency in all PC hardware.

Not only PC hardware has this issue nearly all advanced electronic devices severely lose efficiency past 60C and shuts down around 83 to 85C.

In my experience this is common in smartphones portable computers of all types music amplifiers home car and concert audio types.

Electronic devices deal with heat less efficiently than mechanical devices primarily due to their component materials 

Mechanical devices are general made of iron based alloys and pressure treated. Such materials have higher heat tolerance.

Electronic devices are based around silicon molds that provide pathways for metals with high conductivity but low thermal tolerances like copper silver gold platinum(in rare cases) even mercury and zinc. The last 2 are fairly outdated and have been mostly replaced due to safety concerns or better more effective materials have replaced them.

Feel free to let your fans scream when gaming. There purpose is to dissipate heat don't hold those puppies back. Upon the door let them run around the yard and burn off that energy. 

[Geforsikus_2021]

Hi, if the video card is under warranty, return it under warranty and take another one while the card is working, you never know if it will completely fail with such problems, then it will be problematic to prove that you only played games on it, and not mining, of course, as an option 100% working to find out what exactly causes these problems, it is either to have another video card obviously working and try to play the same games and programs on it, or go with the video card to a friend or acquaintances and check with them, and it is desirable that the power supply was at least 500 watts, but it is better to check with yourself, since if there is a problem with the power supply, then there will be similar symptoms on another video card in the form of a bsod, and so on, you can, as an option, measure the voltage using a multimeter on the power supplies on the 12v line, or look in the aida 64 program (computer tab => sensors) and watch the voltage on the 12v line in a simple PC as soon as they turned on nothing to run, there were similar problems during repair on the Daniil Gerasimov video channel I can't tell you, I don't remember what it was called there, the problem was in the power supply, it simply gave out too high voltage instead of 12v, it gave out almost 13v, and therefore there were problems, problems with replacing the power supply went away. So if there is another video card or another power supply, try both 100% to find the source of the problem, and then decide what to do if the video card, then you will return it under warranty, and if the power supply then you will need a new one.

[Krystallize]

Alright so after actually opening my case I realized that what seemed to me as 2 cables were actually one cable...so sorry for that. I exchanged the cable now with 2 separate cables plugged into PCIe 1 and 3 and pray that it'll work.

I'll give you feedback once some time passes but I'm positive that it will solve my issues, thanks a lot to everyone and sorry again for not checking properly

[amdman]

PSUs and how to configure them is getting more and more complicated. I hope you get it fixed, I have a lot of information and links saved for when I see a problem like yours. It's difficult. I hope this fixes it for you!

[Krystallize]

Just a small update, I've been playing some Halo and CP2077.

Didn't encounter a crash in CP2077

Halo still crashes but different from before, it's only the game and not the drivers, so a pretty good sign

[MUSKEEE]

I JUST PURCHASED RADEON 6900XT REFERENCE CARD TOO AND HAD CRASHING ON THE START OF ALL MY GAMES FOR SOME REASON EVEN WITH THE LATEST DRIVERS. IT WAS A GOOD GUESS REMOVING 2 OF THE 4 MEMORY STICKS AND IS WORKING FINE NOW WITH A DOZEN HOURS OF GAMING WITH SEVERAL DIFFERENT TITLES. MY EQUIPMENT WAS WITH 4 GSKILL CL 14 @ 3200 OF RAM  8 GIGS EACH. I BET IT DOESNT LIKE SINGLE SIDED RAM? 2 STICKS ARE OK. BTW IM USING A 650W EVGA POWER SUPPLY , WITH NO PROBLEMS... GOOD LUCK

[MUSKEEE]

I HAD TO DIVE BACK INTO THIS 6900XT CRASHING PROBLEM AS PREVIOUSLY MENTIONED. MY NEW 6900XT REFERENCE VIDEO CARD CRASHED ONLY WITH 4 STICKS OF MEMORY, 2 WORKED FINE AT 8G A PIECE. WELL I INSTALLED THE LATEST MB CHIPSET DRIVERS AND THE PROBLEM DISAPPEARED ALL GOOD NOW. ASUS RECCOMENDS OLDER DRIVERS ON THEIR SITE ENDING IN .706 WHILE AMD RECCOMENDS THE .431 DRIVERS WHICH CAME AFTER... MY MACHINE HAS A B450-F WITH A 5800X CPU, 4 GSKILL CL 14 RAM STICKS AT 32G TOTAL AND A 650 EVGA POWER SUPPLY.  I ALSO TRIED 2 GSKILL NEO 3600 STICKS AT A 16G TOTAL AND THEY WORKED FINE TOO BUT THE CL 14 STICKS WERE SLIGHTLY BETTER PERFORMERS. THE 6900XT IS MY MOST POWERFUL VIDEO CARD AND JUST FOR AN EXAMPLE TIME SPY SCORES AT 18092 THATS ROUGHLY 7000 POINTS BETTER THAN MY NVIDIA 2080.