RTX 3090 black screen when coming down from a load

[ElMurt]

Hi all!

 

A couple months back I upgraded from a GTX 1080 to a RTX 3090 that I bought second hand. Before switching to the new graphics card the system was running completely fine without issues. After switching I've been chasing an issue where the screen goes black (no even illuminated, just no signal) within a short period after the GPU going to a lower load (exiting game, stopping a benchmark). Any Furmark benchmark and such run complete without issue, even when stressing the CPU as well at the same time. Temperatures seem completely okay (max. 77 degrees celsius on the GPU).

 

What I've tested so far (writing part of this from memory, as I've done some of these things months back, this is not the definitive list of what I've tried):

- With the small form factor case, there comes a riser included. I've read that I have a lower PCI-E version, so I've turned PCI-E to that lower version in the BIOS.

- Updated the GPU with new firmware using the REBAR tool from Inno3D to also enable resizable BAR.

- To make sure the cooling of the CPU is not the problem, turned off turbo boost as the small cooler can't dissipate enough heat otherwise.

- Ran the MATS test (./mats -b 10 -e 30) on the GPU, no issue at all, differing sizes from 5MB to 50MB, all a pass.

- Ran the vulkan_memtest tool for close to two hours. No issue and all passes, but upon closing the tool using CTRL+C it immediately did the blank screen crash.

- Reinstalled Windows 11 completely from scratch after a while with the new card in the system.

 

My hardware:

Case: Fractal Design Ridge Black

Motherboard: MSI MEG Z690i mini-ITX

PSU: Corsair SF750 (2018)

GPU: Inno3D iChill X4 RTX 3090 24GB

SSD: Samsung 980 Pro m.2 1TB

CPU: Intel i7-13700K

RAM: 64GB

 

I feel a bit out of my depth as to what to try next. My hope was it's one memory channel that has issues that I can hopefully turn off so that it'll run with 20GB instead of 24GB, which is fine by me.

 

If any of you have any suggestions that I should try, please reach out. Thanks a lot in advance!

[strange13930]

1 minute ago, ElMurt said:

Hi all!

 

A couple months back I upgraded from a GTX 1080 to a RTX 3090 that I bought second hand. Before switching to the new graphics card the system was running completely fine without issues. After switching I've been chasing an issue where the screen goes black (no even illuminated, just no signal) within a short period after the GPU going to a lower load (exiting game, stopping a benchmark). Any Furmark benchmark and such run complete without issue, even when stressing the CPU as well at the same time. Temperatures seem completely okay (max. 77 degrees celsius on the GPU).

 

What I've tested so far (writing part of this from memory, as I've done some of these things months back, this is not the definitive list of what I've tried):

- With the small form factor case, there comes a riser included. I've read that I have a lower PCI-E version, so I've turned PCI-E to that lower version in the BIOS.

- Updated the GPU with new firmware using the REBAR tool from Inno3D to also enable resizable BAR.

- To make sure the cooling of the CPU is not the problem, turned off turbo boost as the small cooler can't dissipate enough heat otherwise.

- Ran the MATS test (./mats -b 10 -e 30) on the GPU, no issue at all, differing sizes from 5MB to 50MB, all a pass.

- Ran the vulkan_memtest tool for close to two hours. No issue and all passes, but upon closing the tool using CTRL+C it immediately did the blank screen crash.

 

My hardware:

Case: Fractal Design Ridge Black

Motherboard: MSI MEG Z690i mini-ITX

PSU: Corsair SF750 (2018)

GPU: Inno3D iChill X4 RTX 3090 24GB

SSD: Samsung 980 Pro m.2 1TB

CPU: Intel i7-13700K

RAM: 64GB

 

I feel a bit out of my depth as to what to try next. My hope was it's one memory channel that has issues that I can hopefully turn off so that it'll run with 20GB instead of 24GB, which is fine by me.

 

If any of you have any suggestions that I should try, please reach out. Thanks a lot in advance!

Did you DDU the old drivers and download the ones for the 3090

[ElMurt]

1 minute ago, strange13930 said:

Did you DDU the old drivers and download the ones for the 3090

Yes, sorry for not stating that, I reinstalled Windows 10 completely from scratch as one of my solutions as well.

[strange13930]

1 hour ago, ElMurt said:

Yes, sorry for not stating that, I reinstalled Windows 10 completely from scratch as one of my solutions as well.

Have you tried windows 11?

[ElMurt]

53 minutes ago, strange13930 said:

Have you tried windows 11?

Sorry, that was a typo from me. I daily drive a Mac so got all the Windows versions mixed up. It is Windows 11.

[leclod]

One thing I'd try, is without the riser cable. If need be take the system out of the case on plain cardboard.

(I guess you updated the bios of that mobo)

I'd undervolt that 3090. If power is the problem, using less power might help. If heat is the problem, generating less heat might help.

(By the way, I bet you could also undervolt that cpu easily as I do using the MSI cpu lite load mode. I switched from mode 18 down to 11.)

Edited April 18 by leclod

[ElMurt]

I just figured out that running vulkan memtest and then quitting triggers the crash so far 4/4 times. I'm going to disassemble the PC and put it on a cardboard workbench to test and see if it still has that. I'll report back, thanks for the replies so far!

[ElMurt]

2 hours ago, leclod said:

One thing I'd try, is without the riser cable. If need be take the system out of the case on plain cardboard.

(I guess you updated the bios of that mobo)

I'd undervolt that 3090. If power is the problem, using less power might help. If heat is the problem, generating less heat might help.

(By the way, I bet you could also undervolt that cpu easily as I do using the MSI cpu lite load mode. I switched from mode 18 down to 11.)

Unfortunately I just experienced a crash again while have the card directly into the PCI-E slot on the motherboard. It didn't crash on the vulkan memtest so I was completely ready to accept it as a victory and e-mail Fractal Design to request a newer PCI-E 4.0 riser for the case.

 

After shutting down a combination of Furmark and the vulkan memtest the PC operated normally, but while typing that request, it crashed again on me. Unfortunately not the fix then.

 

Does anyone have any more suggestions, or is the GPU probably toast? I want to stress that it doesn't happen under load, quite the opposite actually, immediately or shortly after not using the GPU intensively anymore.

[strange13930]

1 minute ago, ElMurt said:

Unfortunately I just experienced a crash again while have the card directly into the PCI-E slot on the motherboard. It didn't crash on the vulkan memtest so I was completely ready to accept it as a victory and e-mail Fractal Design to request a newer PCI-E 4.0 riser for the case.

 

After shutting down a combination of Furmark and the vulkan memtest the PC operated normally, but while typing that request, it crashed again on me. Unfortunately not the fix then.

 

Does anyone have any more suggestions, or is the GPU probably toast? I want to stress that it doesn't happen under load, quite the opposite actually, immediately or shortly after not using the GPU intensively anymore.

Do you, family, or a friend have a different pc you can test it in?

[ElMurt]

1 hour ago, strange13930 said:

Do you, family, or a friend have a different pc you can test it in?

I've just tested it in an old build I had laying around. I already thought the GPU was the issue, but now confirmed it. I see the same behaviour on a fresh install on the other machine.

 

Thanks for your replies!

[strange13930]

8 minutes ago, ElMurt said:

I've just tested it in an old build I had laying around. I already thought the GPU was the issue, but now confirmed it. I see the same behaviour on a fresh install on the other machine.

 

Thanks for your replies!

Sorry that it had to be the Gpu that was the problem.

[ElMurt]

The GPU has been replaced by a RTX 5080 now and that has solved the issue, haven't had a crash since. Thanks for all the help to both of you!

[AddyP]

Hi all,

came here to look for a solution but now want to leave a guess/ share my experience. It could be the drivers. Had the exact same problem and came to the same solution/ narrowed it down to psi or GPU. At last the drivers were to blame! Current hotfix 576.15 reduced the black screen to black flickering. Before you thrash/ sell your GPU for cheap trie giving it a shot.

 

Kind regards 

Addy

[ElMurt]

Hi Addy,

 

Only thing I could find in the changelogs for the current update is that a blank screen issue is solved with regards to Multiplane Overlay operations, which I think is only a thing on Windows 10, but I could be wrong.

 

Anyways, seeing as I've had crashes on both Linux and Windows with the old GPU and with or without driver, I think it's a moot point in my case. However I'm glad you found a solution for yourself!

 

I've given my old GPU to a local repair business that tries to repair it and use it for PC's for children in poverty that a friend of mine works at. One hell of a GPU they might get if they manage to fix it .

 

If I hear back about it, I can update here.

[AddyP]

Aaah, I missed the part without drivers. But your solution seems great, I hope they can fix it!