Question Random reboots after switching GPU to a 6900 XT ?

[Lumium]

Hello. I'm facing random PC reboots ~0-2 times per day.
It can happen during random YouTube videos or in games. Unfortunately, I can't catch any specific way to achieve it.
Sometimes my cursor and keyboard start to be laggy for 5 - 10 seconds. It is also a rare case (0 - 2 times per day).

In Event Viewer it always:
Event id: 41
Bugcheck code: 0
The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly.

It started after I've switched from Nvidia 980ti to AMD 6900xt + BIOS update.
With 980ti and old bios PC was absolutely stable. After the BIOS update, I set absolutely the same overclock settings.

What I already did:

• Uninstalled all drivers via DDU from safe mode
• Uninstalled all Realtek drivers (network, sound, etc)
• Replaced cables for GPU power
• Rebuilt my PC and checked all cables
• Executed stress tests for GPU only, CPU only, and both at the same time
• Used memtest
• Checked S.M.A.R.T for both disks

My system was stable without any BSOD or reboots for ~1 уear. It starts happens only after installing 6900xt.
What do you think about it? What and how can I test?

HWMonitor during ~1 hour GPU and CPU stress tests:

https://imgur.com/a/EfOMqPM

View: https://imgur.com/a/EfOMqPM


Strange spikes during AMD Adrenaline stress test:

https://imgur.com/a/FvltbBY

View: https://imgur.com/a/FvltbBY

During OCCT and FurMark stress tests I don't have any clocks/power/etc spikes

---
CPU: AMD Ryzen 5 3600 6-Core Processor (oveclocked to 4.20 at 1.35v)
CPU fan: DARK ROCK 4
Motherboard: MSI B450 TOMAHAWK MAX (MS-7C02) with latest bios
Ram: Patriot Viper Steel [PVS416G360C7K] 8x2 with XMP profile (3600 17cl)
SSD/HDD: Samsung SSD 850 EVO 250GB for windows and SiliconPower P34A80 [SP001TBP34A80M28] for other
GPU: AMD Radeon RX 6900 XT Red Devil [AXRX 6900XT 16GBD6-3DHE/OC] with SAM enabled
PSU: Fractal Design Ion+ 860 Platinum
Chassis: I don't remember. It's big mid tower with 140mm fans
OS: Microsoft Windows 10 (10.0) Professional 64-bit (Build 18363)
Monitor: ASUS TUF Gaming VG27AQ 27

 

[Metal Messiah.]

A random reboot mostly tells us that the system has an unstable or faulty/under-powered PSU, but you are rocking a pretty decent PSU nonetheless. There is no hard and fast rule to this though, as some other component might also be the culprit here.

Do you have a spare PSU to test the graphics card ? The Kernel-Power Event ID 41 critical error is generated when a Windows computer starts after its last shutdown and Windows finds that the computer was not previously shut down cleanly.

Event ID 41 reports that something unexpected happened, thus preventing Windows from shutting down correctly.

EDIT: How OLD is your current power supply unit ?

 

[Lumium]

Do you have a spare PSU to test the graphics card ?

I have old Corsair cx600m Bronze, but I'm very unsure that it will be able to handle 6900xt.

EDIT: How OLD is your current power supply unit ?

I've used fractal PSU for ~1 year with 980Ti and it works perfectly. So around 1 year.

Also, I'm thinking about using separate 3 cables for GPU. Now I'm using 8+8 and 8.
Can it help here?

 

[Metal Messiah.]

If your GPU requires 3 "8-PIN PEG/PCI" connectors/cables, then your current PSU can enough cables to power this card. It has 6 "6 +2" PCIe cables for any card. Just make sure all the 3 8-pin cables are properly plugged in the GPU from the power supply. What do you mean by separate cables ?

And yeah, kindly avoid using that Corsair CXM PSU if possible, because I think the problem might be elsewhere instead, imo.

 

[Lumium]

If your GPU requires 3 "8-PIN PEG/PCI" connectors/cables, then your current PSU can enough cables to power this card. It has 6 "6 +2" PCIe cables for any card. Just make sure all the 3 8-pin cables are properly plugged in the GPU from the power supply. What do you mean by separate cables ?

Yes, they plugged properly.
I mean, that now I'm using 2 cables. Each cable has two 6+2 connectors. So the first uses both 6+2 connectors and the second only one 6+2.
And I want to connect 3 cables and use only one 6+2 connector for each. I'm very unsure that it can change something.

https://imgur.com/a/HVqz9n7

View: https://imgur.com/a/HVqz9n7

 

[Metal Messiah.]

Nope, I really don't think using separate cables is gonna solve the problem imo.

 

[Lumium]

Nothing changed a all, but the PC worked stable for ~2 days. It was stable even under heavy load.
Today I left it AFK without any heavy app (just browser). And it rebooted with a power error.
So I'm hardly suspecting my PSU or home electricity. I'm very unsure that a faulty GPU can cause a reboot without any load. Especially, when it can handle stress tests without issues.

Also, can BIOS cause this? I want to check all possible software solutions at first.

 

[RodroX]

Nothing changed a all, but the PC worked stable for ~2 days. It was stable even under heavy load.
Today I left it AFK without any heavy app (just browser). And it rebooted with a power error.
So I'm hardly suspecting my PSU or home electricity. I'm very unsure that a faulty GPU can cause a reboot without any load. Especially, when it can handle stress tests without issues.

Also, can BIOS cause this? I want to check all possible software solutions at first.

1. Are you using 3 different cables to plug the GPU?, this mean 3 different cables coming from your PSU.
2. Did you made a restore default setting in BIOS (+ Save and Exit), before and after you done the bios update?
3. Are you avoding the lastest windows update for any reason (You are runing the 2019 november update) ?
4. When was the last time you updated AMD chipset drivers (https://www.amd.com/en/support) ?

5. As a last resort, I would: Backup all your important user files (documents, images, videos, etc), get the newest Windows 10 version (Download Tool: https://www.microsoft.com/en-us/software-download/windows10), make a fresh clean windows install, download all the lastest driver from AMD, install chipset first, then VGA, make sure you update network and sounds drivers too (either form MSI website or using the device panel in windows to check for new drivers) and finally run windows update for new patches.

 

[Lumium]

Are you using 3 different cables to plug the GPU?, this mean 3 different cables coming from your PSU.

No. For now, I'm using 2 cables. As I mentioned above.
I think I should have constant reboots under stress tests if GPU doesn't have enough power. Am I right?

Did you made a restore default setting in BIOS (+ Save and Exit), before and after you done the bios update?

No, I've updated bios without resetting settings. But I've fully reset it after the update via MB battery.

Are you avoding the lastest windows update for any reason (You are runing the 2019 november update) ?

I think, Build Number is related to the used build during installation. I've fresh updates too.

https://imgur.com/a/vVxSW1q

View: https://imgur.com/a/vVxSW1q

When was the last time you updated AMD chipset drivers (https://www.amd.com/en/support) ?

I've reinstalled them after DDU Nvidia drivers.


UPDATED: Yes, I'm thinking about reinstalled windows too. But this error doesn't look like a software error. I think I should have some error in the log or a simple BSOD at least.

 

[RodroX]

I would use 3 different cables for such power hungry GPU anyways.
I would have my Windows 10 up to date (specially when using new hardware). Im using Home edition and runing version 21H1 (19043.1151).
I would keep my chipset drivers up to date. DDU won't uninstall chipset drivers far as I know. What you did there was remove GPU drivers and install new ones. Chipset drvivers are a very different thing which control many functions of your mobo, including you R5 3600 power management.
If everything fails and you don't wana make a new clean install of windows I would try a reset defaults settings + Save and Exit in BIOS anyways.

 

[Lumium]

DDU won't uninstall chipset drivers far as I know.

Yes, of course. I meant, that I removed Chipset drivers at the same session with cleaning AMD drivers.

If everything fails and you don't wana make a new clean install of windows I would try a reset defaults settings + Save and Exit in BIOS anyways.

I've installed new clean and licensed Windows 10, reset BIOS to it defaults and use 3 cables.
So let's wait a bit.

Also, I'm able to reproduce another trouble. Always in the same place in Cyberpunk I got my screen black and sound lost.
But my CPU fan continues to work, GPU fans not working, but highlight works. I'm still can't catch this during any stress tests or other locations, only in concrete place in Cyberpunk.
I've tested it with low settings (~180watt for GPU at max), stock configuration for RAM/CPU, and fresh windows.
Is it something related to GPU drivers?

Update During this bug I have ~60% CPU load with 55 temp, and GPU 90% with 70 temp.
Update 2 Wow. I have 2 monitors (1080 usually off and 1440), so I decided to check the game on the 1080 display. And I was able to pass the crashing moment.
I've switched back to a 1440p monitor, set 1080p/Fullscreen, and got a PC crash. I've tried it a few times and always have the same result at any game setting.
So I have replaced the cables from 1080p to 1440p and was able to pass the crashing moment. Now I'm testing it with my old overclocking options and seems like it works.
I don't believe that a new cable can change anything, but I can't achieve the same crash even after ~10-15 tries. Look's like temporary luck with some faulty hardware, so I'll continue my tests.

 

[RodroX]

Yes, of course. I meant, that I removed Chipset drivers at the same session with cleaning AMD drivers.


I've installed new clean and licensed Windows 10, reset BIOS to it defaults and use 3 cables.
So let's wait a bit.

Also, I'm able to reproduce another trouble. Always in the same place in Cyberpunk I got my screen black and sound lost.
But my CPU fan continues to work, GPU fans not working, but highlight works. I'm still can't catch this during any stress tests or other locations, only in concrete place in Cyberpunk.
I've tested it with low settings (~180watt for GPU at max), stock configuration for RAM/CPU, and fresh windows.
Is it something related to GPU drivers?

Updated During this bug I have ~60% CPU load with 55 temp, and GPU 90% with 70 temp.

First of all I wana be clear, Im not a GPU or PSU expert (I do have more than 28 years of experience building, repairing and gaming on PC. I will try to keep it clear and simple (specially since english is not my natural lnaguage)


It seems to me Cyberpunk is the game that can really stress GPUs. And it also seems that the RX 6900XT and RTX 3080TI and 3090 are the ones that suffer the most issues. But thats not the end of the story.

Is this because of the GPU, a driver, or the game? Perhaps is a mix of a few things.

What I think is that both AMD and Nvidia have pushed the performance of their silicon to the very edge (trying to fight each other). In the high end GPU range (your GPU, and the RTX 3080/TI and 3090) this mean push them right into the "big and unexpected problems" zone. What I mean with this, is that for me pushing this GPU soo hard have probably made them more susceptible to failures, specially from bad or not decent enough power delivering and/or cooling.

This GPU can push the power supplies to their limit too. They can have spikes of power really, really high in a matter of micro seconds, and if the PSU can't deliver soo much power soo fast it can end up in strange behaviors. And this is not the first time in PC history that the PSU were not "ready" for a new generation of GPUs.

I guess all you have left to do its test and see if the issue is only with Cyberpunk.

Then again this is my personal opinion, from what I have tested, read, and helped clients, friends and family that have the money to buy this new monsters.

 

[TK31]

Yes, of course. I meant, that I removed Chipset drivers at the same session with cleaning AMD drivers.


I've installed new clean and licensed Windows 10, reset BIOS to it defaults and use 3 cables.
So let's wait a bit.

Also, I'm able to reproduce another trouble. Always in the same place in Cyberpunk I got my screen black and sound lost.
But my CPU fan continues to work, GPU fans not working, but highlight works. I'm still can't catch this during any stress tests or other locations, only in concrete place in Cyberpunk.
I've tested it with low settings (~180watt for GPU at max), stock configuration for RAM/CPU, and fresh windows.
Is it something related to GPU drivers?

Update During this bug I have ~60% CPU load with 55 temp, and GPU 90% with 70 temp.
Update 2 Wow. I have 2 monitors (1080 usually off and 1440), so I decided to check the game on the 1080 display. And I was able to pass the crashing moment.
I've switched back to a 1440p monitor, set 1080p/Fullscreen, and got a PC crash. I've tried it a few times and always have the same result at any game setting.
So I have replaced the cables from 1080p to 1440p and was able to pass the crashing moment. Now I'm testing it with my old overclocking options and seems like it works.
I don't believe that a new cable can change anything, but I can't achieve the same crash even after ~10-15 tries. Look's like temporary luck with some faulty hardware, so I'll continue my tests.

I have a 6900XT as well and faced a similar issue with Cyberpunk. In the wastelands area it was crashing for no reason, and it would kill the AMD Drivers to the extent where every game you played after it would continually crash the driver unless you restarted the PC. It was solved for me to some extent by turning off Ray Tracing. Looks like Cyberpunk is just poorly optimized for AMD.

In regards to your power question, I had a similar issue previously and i had a Corsair AX860 80+Platinum. It would continually shut off without warning. Turned out to be the PSU in my case as I had a spare PSU to test it out and a replacement sorted out the problem. I would highly recommend testing with a spare PSU if you have one especially if your PSU is older.

 

[Lumium]

And this is not the first time in PC history that the PSU were not "ready" for a new generation of GPUs.

Yeah, I though that new 860 platinum will be enough. I'm expected to take 3070ti, but according to benchmarks 6900xt fits extremely well for 1440. Especially, while it costs just 15 - 20% more, than 3070ti.

i had a Corsair AX860 80+Platinum. It would continually shut off without warning

Looks very similar to my case.


So looks like I've left only one reasonable option to buy and test with the new PSU. I hope plat/gold 1000+ should be more that enough.

 

[RodroX]

Yeah, I though that new 860 platinum will be enough. I'm expected to take 3070ti, but according to benchmarks 6900xt fits extremely well for 1440. Especially, while it costs just 15 - 20% more, than 3070ti.


Looks very similar to my case.


So looks like I've left only one reasonable option to buy and test with the new PSU. I hope plat/gold 1000+ should be more that enough.

I would aim for some of the Tier A PSU (https://forums.tomshardware.com/threads/psucultists-psu-tier-list-rev-14-8-12-07-2021.3624094/) (https://linustechtips.com/topic/1116640-psu-tier-list/), and make sure it does not have any reported unwated issues.

I hope you get a good unit and so the issue gets fixed.

 

[Lumium]

Looks like I have only one choice not out of stock: ASUS ROG-STRIX-1000G
But I'm not sure about 1 rail PSU for 6900xt.

Also, looks like my current PSU in A tier already

 

[TK31]

Looks like I have only one choice not out of stock: ASUS ROG-STRIX-1000G
But I'm not sure about 1 rail PSU for 6900xt.

Also, looks like my current PSU in A tier already

I switched to a Corsair RMx 1000W and it solved the shutdown issue for me

 

[Lumium]

A bit of necroposting.
I finally got Asus Thor 1200W (instead of Fractal), and my PC rebooted after ~10 minutes of use.
Same event - 41 Kernel-Power.

Looks like my last hope is to roll back the BIOS and/or test the GPU with a different CPU/MB/RAM.

P.S. Asus Thor is kinda strange PSU. It costs more than twice compared with Fractal, but all cables and slots feels like my super cheap noname PSU from Aliexpress. Let's see how long and stable it will work.

 

[Hashman57]

What do you mean "let's see how long it will work"?
You said that it "rebooted after ~10 minutes of use".

 

[Lumium]

What do you mean "let's see how long it will work"?
You said that it "rebooted after ~10 minutes of use".

It was about the new PSU. I've tested it with the 3090 GPU. And looks like PSU is not an issue here.
My PC with 6900xt still have random reboots.

 

[hanara93]

Hey,
@Lumium did you find any soulution?
After replacing my RTX 2070Super into RX6900XT Power Color I have same issue, my computer randomly freezing - black screen, PC restart. Never happen before on Nvidia card. What do I do?

 

[InvalidError]

If you change hardware and firmware, overclock settings that used to work may not necessarily still work. First thing you should do is go back to stock and see if that makes the problem go away.

When you overclock, you eat into error margins that are normally there to ensure everything works under circumstances. You changed the GPU, now the power delivery system including interactions between the PSU, motherboard and GPU may not be behaving in a way that is compatible with your old OC anymore.

 

[Lumium]

Hey,
@Lumium did you find any soulution?
After replacing my RTX 2070Super into RX6900XT Power Color I have same issue, my computer randomly freezing - black screen, PC restart. Never happen before on Nvidia card. What do I do?

Hello. No, I still suffer from this problem. It is pretty rare, maybe 1 - 2 times a month.
But It is still often for ~20 - 30 seconds after PC power up. Looks like using sleep instead of shutdown solving it.

If you change hardware and firmware, overclock settings that used to work may not necessarily still work. First thing you should do is go back to stock and see if that makes the problem go away.

When you overclock, you eat into error margins that are normally there to ensure everything works under circumstances. You changed the GPU, now the power delivery system including interactions between the PSU, motherboard and GPU may not be behaving in a way that is compatible with your old OC anymore.

Yeah, I've tested it with the full BIOS reset. I've used it for 1 month in stock mode, and I don't see any difference.
I've run different stress-test for hours, and all passed successfully.

So maybe I just unlucky with the quality of my GPU.

 

[Brandon.H]

It’s unfortunate we’re having these issues however it’s good to know I’m not alone. I’m running an ASUS ROG STRIX B-550F, Corsair RGB 32gb 3600 ddr4, AMD Ryzen 5900X, EVGA 1000W Gold rated PSU, and AMD RX6900XT ULtimate Red Devil. I can play Tarkov with 0 issue. Rainbow six was causing my ox to reset every 2 minutes into a match.(turning VSYCNH on within the game 100% fixed this) sea of thieves, 5 minutes to 2 hours and it causes my ox to reset. Same with ARK: Survival evolved. Neither of these have an in-game option for me to turn on V-Synch. And limiting the FPS via the Radeon app is actually limiting my FPS. I’ve noticed this is happening in games where I’m getting 300-800FPS. Sound similar?

 

[Lumium]

Hmm... To be honest I don't have any troubles with different GPU or CPU loads. I can have unexpected reboots during the simple YouTube video or heavy games.
But unfortunately, I've not tried to play any of these games. I'm playing Cyberpunk, Apex, Expedition: Rome, Minecraft, and Elden Ring.
I have FPS troubles with the old games, but as I know, it is expected behavior for RDNA2.

Looks like it is a driver issue because I have fewer reboots than just after buying 6900xt. I had ~0 - 2 reboots per day at the start. Now it is ~1 - 2 reboots per month.

I don't want to blame AMD, but this is a bit strange. I've checked with my old GPU (980ti), and I didn't have any reboots for ~1 month. Maybe just luck, who knows.