Blue Screen of Death in Windows

I think BMD recommended not to disable it as Windows would still keep on the list of devices (somewhere) but without driver Resolve would complain that there is an "unsupported GPU" in the system.

Try disabling it by all means but it might not work.

Can intel GPUs be disabled from the BIOS? That might make it fully hidden to the OS.

Have you set the GPUs to manual in Resolve? Basically telling Resolve not to use your iGPU.

Let me rephrase that...

Disable it at the BIOS level so it does not appear in the Device Manager, thus invisible to the OS and all apps.

Ok, guys, I either have narrowed the list of causes to Red Giant or I'm just (un)lucky and the issue reproduced itself once I enabled its process... The thing is I actually use 2 effects from Red Giant pack (Universe). A transition on a text+ layer in each clip in a 9 by 16 projects for social media and a glitch effect from the said pack between the intro (1) and outro (2), so only two this effects in each project. I now am going to uninnstall it and see if it really works. Strangel - been using RG effects and transitions for years in Premiere Pro with no issue, but seems in DR there's a driver-level conflict between SuperScaling and the said transition.
I'll be back with the results...

Spoiler - no luck. I'm out of variants.
To sum it up:
- to reproduce the issue, create an empty timeline (1920x1080), place a video on it, turn on NVidia SuperScale x4, turn on and analyze Lens correction. Place the playhead over the video and move it back and forth. It may happen instantaneously, in 5 seconds or in some minutes
- when it happens, it's always a dxgmms2.sys bsod
- it happens with any material and any codec
- it happens with 570.xx and 566.xx drivers
- it happens with new and older iGPU drivers
- it happens with the iGPU turned off (in the device manager which is = in Bios)
- it happens both with third-party plugins used in the project and without them, even with an unedited video.
- it happens with XMP and CPU downvolting as well as BIOS set to defaults
- it happens with all the non-windows processes turned off on load
- it happens only in Davinci Resolve.
- didn't happen when SuperScaling was not used, but I'd worked in DR for 2 weeks (12 short videos made), so this one is not 100%.

I sent bug reports to:
- Blackmagic - no response
- Nvidia - suggested workarounds none of which helped (DDU, Disable Hardware Accelerated GPU scheduling in Windows, set Nvidia settings to forced 4080 in Cuda, OpenGL and max powerplan, turn off services in MSCONFIG > Services).
- Palit (my card's vendor) - "we don't make drivers, ask NVidia"
- Microsoft - no respose. MS community: "try older drivers"

I'm out of ideas. The universe doesn't want me learning Davinci Resolve.

Screenshot 2025-03-24 204924.png (74.59 KiB) Viewed 4249 times

As previously, I don't get a BSOD but an error.

Sorry but I don't think there is anything you can do here to resolve this issue. This feature does not work with your HW combination or mine.

As previously, I don't get a BSOD but an error.

Sorry but I don't think there is anything you can do here to resolve this issue. This feature does not work with your HW combination or mine.

Funny I've got a BSOD just now, having edited the audio of a clip, then colored the clip itself in Color tab, then trying to switch to Fusion... Didn't use a SuperScale at all or anything in the video tab of the Inspector... Looks like I have to go back to Premiere Pro...

I hate when I am given this option but... have you tried a clean windows installation?

I hate when I am given this option but... have you tried a clean windows installation?

Today's task that is.

Just this morning I was reading about nVidia drivers being pretty bad recently, someone suggests nVidia is focusing too much on the 5000 series.

Good luck with the new install! But I'm pretty confident it won't fix the SuperScale issue. Hopefully it won't cause a BSOD though...

As previously, I don't get a BSOD but an error...

Tony since you can reproduce error 700 (which is a VRAM memory allocation failure) on your Windows machine, can you run that same scenario on your M4 Mac Mini? What happens then?

it just works - though it's not smooth, I have the base mac mini.

That said, when I closed Resolve, macOS complained that Resolve quit unexpectedly. But while I was scrubbing the timeline, no errors.

Was that 4k using 4x Super Scale, or 1080 or what? And what frame rate?

4x super scale. 4K content on FHD timeline. 30fps.

The default project, as on my Windows desktop

Reinstalling Windows did not help, even with a basic set of programs (browser, Shutter Encoder, Adobe pack, no anitivirus, no anything else). 3 days of perfect video editing in DR until on the fourth I enabled SuperScale and immediately got a bsod. I asked a friend to install DR too - he's a Premiere Pro guy, never worked in DR, and once he uses SuperScale, DR either crashes or freezes in a way even closing the process does not work - he had to reboot. Everything points to Win+Nvidia+DR Superscale again. Never gonna use it again until it's fixed somehow...

To me it was altready pretty clear that there is no solution for the SuperScale issue on your end at the moment!I was mainly referring to other instabilities.
Hopefully BMD/nVidia will improve that feature one day.

I use 2x Super Scale to upscale 1080p footage to a UHD timeline when rendering with no problems.
I may be wrong here but to me, using 4x with 4k footage means you are trying to upscale it to 16k which is why your system throws in the towel.

I use 2x Super Scale to upscale 1080p footage to a UHD timeline when rendering with no problems.
I may be wrong here but to me, using 4x with 4k footage means you are trying to upscale it to 16k which is why your system throws in the towel.

I tried FHD materials with the same result. Nonetheless my system is capable of running it as 16K as well with no noticeable frame drops, so it's not the problem

...I tried FHD materials with the same result. Nonetheless my system is capable of running it as 16K as well with no noticeable frame drops, so it's not the problem

I tested Super Scale 4x on 4k material in a 1080p timeline, using my 2017 i7 iMac27, 32GB RAM, 8 GB Radeon Pro 580, on Resolve Studio 18.6.6 and 19.1.4, using both the Metal and OpenCL APIs (in Resolve>Preferences>Memory & GPU). It was really slow (about 0.5 fps), but it ran with no errors.

Have you tried it using OpenCL or just CUDA? That might be useful to know. The error 700 is actually a CUDA error that NVIDIA defines this way: "cudaErrorIllegalAddress = 700: The device encountered a load or store instruction on an invalid memory address. This leaves the process in an inconsistent state and any further CUDA work will return the same error. To continue using CUDA, the process must be terminated and relaunched."

It is conceivable the error 700 might be caused by a CUDA coding error in Resolve itself. However the BSOD is the responsibility of NVIDIA since that is a kernel-mode error.

Looking back through this thread, I don't see anywhere you have uploaded Resolve logs. The first step mentioned on the Support and Troubleshooting FAQ mentions that procedure: viewtopic.php?f=21&t=90190

...I tried FHD materials with the same result. Nonetheless my system is capable of running it as 16K as well with no noticeable frame drops, so it's not the problem

I tested Super Scale 4x on 4k material in a 1080p timeline, using my 2017 i7 iMac27, 32GB RAM, 8 GB Radeon Pro 580, on Resolve Studio 18.6.6 and 19.1.4, using both the Metal and OpenCL APIs (in Resolve>Preferences>Memory & GPU). It was really slow (about 0.5 fps), but it ran with no errors.

Have you tried it using OpenCL or just CUDA? That might be useful to know. The error 700 is actually a CUDA error that NVIDIA defines this way: "cudaErrorIllegalAddress = 700: The device encountered a load or store instruction on an invalid memory address. This leaves the process in an inconsistent state and any further CUDA work will return the same error. To continue using CUDA, the process must be terminated and relaunched."

It is conceivable the error 700 might be caused by a CUDA coding error in Resolve itself. However the BSOD is the responsibility of NVIDIA since that is a kernel-mode error.

Looking back through this thread, I don't see anywhere you have uploaded Resolve logs. The first step mentioned on the Support and Troubleshooting FAQ mentions that procedure: viewtopic.php?f=21&t=90190

Hi, joema4
Tried the OpenCL just now and it seems to run fine though with a tremendous performance drop rendering it unusable. And once I switched the renderer to Auto I got a 700 error this time. It doesn't mean anything good, because once I sweep the cuda folder and rebuild the cuda cache, I'll definitely get a bsod. Tried that so many times.

As for the logs - I created one immediately after this error. I don't think once a bsod happens a log will be of any use because the PC is rebooted and there's nothing to log when anew. But if needed, I'll try to reproduce the bsod and bring a log after it. As I said it doesn't happen instantaneously, you have to try to repo it for some time. The logs: https://drive.google.com/drive/folders/ ... sp=sharing
Really hope it's a cuda-driver issue and it can be fixed.

Thanks for the link by the way. Found some useful info there

[Dwaine Maggart]

OpenCL is not an option for NVIDIA GPUs.

Please go to the NVIDIA driver downloads site and download and install the 566.36 STUDIO driver.

When you run the driver installer, select Custom install, then the Clean install option.

Does that driver help?

There have been many reports of NVIDIA 57x drivers causing BSOD's, so the hope is that reverting to a 56x driver solves the issue.

These days, a BSOD is normally related to a hardware issue. Modern systems, be it Windows or MacOS, are pretty well protected against application software causing it.
Might be a thermal issue or the power supply. DR is hammering hard on the GPU when doing things like SuperScale.

You are correct .. most, if not all BSOD is related to hardware issues becouse as you said modern OS are well protected against applications that may cause it.

These days, a BSOD is normally related to a hardware issue. Modern systems, be it Windows or MacOS, are pretty well protected against application software causing it.
Might be a thermal issue or the power supply. DR is hammering hard on the GPU when doing things like SuperScale.

You are correct .. most, if not all BSOD is related to hardware issues becouse as you said modern OS are well protected against applications that may cause it.

Something wrong with the forum! It duplicates my posts even though I clicked once to submit

These days, a BSOD is normally related to a hardware issue. Modern systems, be it Windows or MacOS, are pretty well protected against application software causing it...

That is by far the most common case. The CPU's hardware memory management unit, as used by x86 and Apple Silicon CPUs on Windows and MacOS, provides near-absolute isolation between different user processes and between user mode and kernel mode.

In a troubleshooting situation, you go from the most likely to the least likely. The chance of a fixable Resolve bug causing a BSOD or Mac kernel panic is *extremely* low.

However, it is not absolutely non-zero, especially on Windows (due to the kernel mode driver situation, e.g, CrowdStrike, and less-safe GPU programming methods). Here is a theoretical scenario: Say a CUDA programming error in Resolve causes the GPU to have an access violation and crash. That is what error 700 implies -- it is not purely a Resolve error message, it is a CUDA error indicating a problem with code executing on board the GPU and within the GPU's local address space. It's similar to an illegal memory reference or access violation in an application program running in Windows or MacOS.

My understanding is when the app (ie Resolve) is compiled, the CUDA kernel is also compiled from CUDA source code in the app to a separate executable. Similar to a separate embedded processor, that compiled CUDA kernel is downloaded and run in the GPU. Any software can have bugs so the CUDA kernel running on the GPU could throw errors or crash.

But even that should not cause a BSOD. Error 700 is a crash or illegal reference in the local VRAM address space of the GPU. It's not in the CPU's kernel mode; it's on a separate processor. But there is also the NVIDIA kernel-mode driver. It may be involved when handling certain error conditions. If it has bugs when handling certain error conditions, it could theoretically cause a BSOD, even if the inception of that sequence started in a user-mode code path in Resolve. I'm not saying that happened in this case, but I think it is theoretically possible.

But you don't start assuming a BSOD is a Resolve error any more than you assume cosmic ray bursts from outer space are causing a few particular machines to crash.

OpenCL is not an option for NVIDIA GPUs.

Please go to the NVIDIA driver downloads site and download and install the 566.36 STUDIO driver.

When you run the driver installer, select Custom install, then the Clean install option.

Does that driver help?

There have been many reports of NVIDIA 57x drivers causing BSOD's, so the hope is that reverting to a 56x driver solves the issue.

Hi, Dwaine, as I told you in the PM, the older drivers didn't help, DR bsods even with 56x drivers, both studio and GR, even on a freshly installed Windows. And it only happens when Superscale is used. Apart from it, I was able to work without any problems even on the current 572.83 Studio driver without using Superscale. Not a single bsod happened within several days. So it all narrows this issue to using Superscale, or - better said - to a combination of SuperScale in Davinci Resolve and something (software or hardware) related to NVidia. I bet it's a software issue. With OpenCL I can use Superscale without a bsod too, by the way, and I can see my dGPU loaded despite it. But for the performance

I can confirm that the issue causing BSOD is with using the AI Super Scale with 4x NVIDIA RTX or 2X NVIDIA RTX. I believe this issue is caused by the latest NVIDIA drivers running on 30 or 40series, for me it happens on my 4090. If anyone on a 50-series can confirm this would be great

The 700 error is basically an access error, normally memory. I too have find the problem in the new release of version 20 of resolve. Seems to be linked to upscaling using the Nvidea cards at 2x and 4x. If I go back to the normal versions it seems OK. I didn’t have this problem with version 19. I get the 700 error and about once in five the blue screen of death. Any solutions yet? Card runs fine with all my Adobe software (RTX 4060)

[Dwaine Maggart]

@Andy and John:

What is the timeline resolution and frame rate, and what is the clip resolution, file type and frame rate of the clip you are applying SuperScale to?

In my case it’s an MP4 file from a mobile. Vertical and at about 600x800. Upscaling to vertical 4K

I can confirm that the issue causing BSOD is with using the AI Super Scale with 4x NVIDIA RTX or 2X NVIDIA RTX. I believe this issue is caused by the latest NVIDIA drivers running on 30 or 40series, for me it happens on my 4090. If anyone on a 50-series can confirm this would be great

I tried all 8 flavors of AI Super Scale in Resolve Studio 20 Beta and no problems:


Also no problems in Resolve Studio 19 so I dont think it is related specifically to 40X and 50X GPUs .. maybe related to some hardware configurations and drivers.

I've been having this issue since September 2024, started on 2 different machines at the exact same time:

Desktop:
Win 11 24H2
64GB RAM
i9-12900K
RTX 3090
6TB SSD space, 2.5 used

Laptop:
ASUS ProArt P16
Win 11 24H2
32GB RAM
AMD Ryzen AI 9 HX 370
RTX 4060 (Laptop)

There was a Windows update, DaVinci Resolve update, and I installed the September 2024 NVIDIA Studio Driver 561.09 all at once. I know, laugh at me, I deserve it.

DaVinci Resolve update wasn't the source, as older versions were having it too

Windows 11 I haven't tried rolling back the update, BECAUSE: Rolling back the NVIDIA Studio Driver to August 2024 (560.81) solved the issue on both machines

I was using DR19.x with the 560.81 Studio Driver for months now, 0 issues on either my 3090 Desktop nor 4060 Laptop, but then DR20 required CUDA 12.8, which meant a 2025 Driver was finally unavoidable.

I used DDU (Display Driver Uninstaller, the best way to do a 100% clean install) to test every NVIDIA driver from 2025 (Jan->March), and strangely enough, the March 2025 GAMING Driver was the most stable.

Only on final renders did it start Error 700'ing / (rarely) BSOD'ing - trying to watch a lot of the timeline quickly caused it too, but I don't do that while editing so it wasn't an issue.

I think there's a VRAM memory leak (from NVIDIA Drivers), as VRAM slowly "Fills up" over hours of editing, until either it throws a "GPU Error 700" at me, or silently crashes. I restart my computer if this happens, and everything is fine.

Since the only issue was exporting, I just turned on Render Cache when ready to render, and then periodically restart my computer every X minutes when VRAM tops out and it throws Error 700. After a few cycles of this, I check "Use Render Cache" in the export settings, and it renders out fine (though sometimes I need to fix corrupted images around the areas where it throws the Error 700)

I noticed the time it took to crash wasn't linear, as some areas could only go 1-2 minutes before crashing, while other areas it got >10 minutes through before having an error.

After trying to isolate the cause, I'm 99% sure it's all the 2x/4x NVIDIA SuperScale images I use. I shoot in 4k, but often use assets which were made for a 1080p world, so that leads to a lot of SuperScale usage.

I hope NVIDIA fixes their SuperScaling soon, as this issue has been going on for approaching 1 year now, across 2 different machines.

@Andy and John:

What is the timeline resolution and frame rate, and what is the clip resolution, file type and frame rate of the clip you are applying SuperScale to?

I work with 1080p timelines, and I use SuperScale with 1080p materials zoomed up so the SuperScale makes a such media a better quality.

I can confirm that the issue causing BSOD is with using the AI Super Scale with 4x NVIDIA RTX or 2X NVIDIA RTX. I believe this issue is caused by the latest NVIDIA drivers running on 30 or 40series, for me it happens on my 4090. If anyone on a 50-series can confirm this would be great

I tried all 8 flavors of AI Super Scale in Resolve Studio 20 Beta and no problems:


Also no problems in Resolve Studio 19 so I dont think it is related specifically to 40X and 50X GPUs .. maybe related to some hardware configurations and drivers.

There will always be a person who's got everything alright. But you see, we have BSODs, there IS a problem and it's not a hardware configuration problem - or it would have happened elsewhere, but it only happens in Davinci Resolve and on NVidia cards\drivers. A lot of people speak of the same problem here, it this topic, and here: https://forum.blackmagicdesign.com/viewtopic.php?f=21&t=215190 And there are even more on Reddit and other comminities, for instance here's a spanish guy who contacted me in PM:

image_2025-04-16_083143027.png (64.27 KiB) Viewed 2803 times

Here's one on NVidia forums: https://forums.developer.nvidia.com/t/b ... gpu/330106

Tom's Hardware: https://forums.tomshardware.com/threads ... d.3877819/

And there are many others. They all have one thing in common - they are new, exactly when the new drivers for NVidia cards came out. And we, as users, cannot make the big company NVidia is check all the cases and roll out a solution. But Blackmagic can.

Hear hear

Andy, I am not trying to dismiss the idea that the problem doesn't exist, surely you and many others are suffering from this problem. But are most users with nVidia 50x and 40x GPUs having this problem? I don't know for sure but my guess is no, otherwise you will see hundreds of related threads in the form. That is why I think this problem could be related to specific hardware configuration/drivers, and not with nVidia 50x and 40x GPUs in general. Another possibility is a faulty hardware.

I hope BMD can help you fix the problem somehow.

Hi, everyone. Just been testing out my footage in two different projects with yesterday's new NVidia Studio drivers. I'm actually running SuperScale - Nvidia 4x Ultra in the background while I'm typing this.
Don't wanna be too optimistic about it but in half an hour I've had zero BSODS... Please try the new driver out too to see if it helps. I hope it will

DVR crashes when I select some fusion effects. Got the latest drivers, haven't a clue why

Ric Marty

DVR crashes when I select some fusion effects. Got the latest drivers, haven't a clue why

Ric Marty

Could totally reproduce the BSOD just now... I thought I had cope with it, gues no - the issue is still there. Two weeks of perfect performance and once I activated the SuperScale in DVR after about half an hour I got a bsod...

Sent all the logs to Blackmagic tech support. Hope they'll understand why this happens and fix it...

To reproduce it, open any project with plenty of materials on the timeline, activate the SuperScale feature on all of them (Nvidia 4x, Ultra), press play, once it's played back till the end - if no bsod happened - start messing around with the playhead.

I feel your pain but I have a question: SuperScale is clearly bugged (on DVR or nVidia's end) - why keep using it?

I'm not saying it's ok for SS not to work -but in the end if it doesn't work, it doesn't work. or not?

I feel your pain but I have a question: SuperScale is clearly bugged (on DVR or nVidia's end) - why keep using it?

I'm not saying it's ok for SS not to work -but in the end if it doesn't work, it doesn't work. or not?

I don't use it, but you know, it just gives me that feeling, when you have a house and from one side the foundation starts to destruct... If there's one thing that doesn't work, what other issues may appear during a serious project?

I stopped using the SuperScale at all, though it's clearly a useful feature that works. But what else can there be in DVR that makes your PC faint?

Many many other things.

Same for Premiere. Windows. macOS. Everything

Many many other things.

Same for Premiere. Windows. macOS. Everything

Premiere (and no other software) never gave me bsods )) That is the catch

yes, BSOD's are rare these days, I can agree on that.

But Premiere gave me many many MANY other things!

I’ve been running into the same issue with the Delta Keyer and Matte Control in Fusion, and it’s even causing the Blue Screen of Death on Windows for me. Glad to know I’m not the only one dealing with this, especially when working on more complex comps. Hopefully, there’s an update or fix coming soon.

I’ve been running into the same issue with the Delta Keyer and Matte Control in Fusion, and it’s even causing the Blue Screen of Death on Windows for me. Glad to know I’m not the only one dealing with this, especially when working on more complex comps. Hopefully, there’s an update or fix coming soon.

As mentioned earlier, a BSOD typically indicates an issue at the Windows or kernel-mode driver level. In the vast majority of cases, these errors aren't caused directly by user-level applications. Usually, the solution involves driver updates from NVIDIA, BIOS updates from your motherboard manufacturer, or possibly adjustments in hardware configuration.

If you'd like to investigate the exact cause of your BSOD, here's a short tutorial that walks you through the steps clearly:

I used ChatGPT to analyse my dump files created after the BSOD had happened last time. Here's what I got:

BSOD ANALYSIS SUMMARY
Bug Check:
SYSTEM_THREAD_EXCEPTION_NOT_HANDLED (0x7E)
NTSTATUS: c0000005 – Access violation (attempted to read from invalid address 0x48)
Culprit:
Module: dxgmms2.sys
Function: dxgmms2!VIDMM_PAGE_DIRECTORY::ExpandLargePagePte
Stack trace points to:
dxgmms2.sys — DirectX Graphics MMS (Microsoft kernel graphics memory manager)
What is dxgmms2.sys?
dxgmms2.sys is a core Windows system driver responsible for managing video memory through DirectX (WDDM). It interfaces with the GPU drivers (NVIDIA, AMD, Intel) and the Direct3D subsystem.

If this module crashes with an access violation (0xC0000005), it usually means:

Bad or unstable GPU driver
Conflict with software rendering or virtualization layers (e.g., Hyper-V, WSLg, RDP, OBS, Parsec, etc.)
GPU instability or overheating (hardware-related in rare cases)
Hyper-V enabled, interfering with GPU functions
Issues in a recent Windows Insider or Preview build
Detailed fault analysis
Exception:
0xC0000005: Access violation — Attempted to read address 0x48
This is a NULL pointer dereference or uninitialized memory access issue. Very likely a software bug inside the memory management component for DirectX (triggered within VIDMM_PAGE_DIRECTORY::ExpandLargePagePte).

Key data from the dump
Faulting module: dxgmms2.sys
Version: 10.0.26100.3912 → This is a Windows Insider Preview build (24H2 series)
Hypervisor: Enabled → IsHyperV=1, AnyHypervisorPresent=1
Process: System (i.e., kernel context)
Conclusion
This BSOD was caused by a read access violation in the DirectX graphics memory subsystem, most commonly due to a faulty or incompatible graphics driver, or GPU virtualization conflicts.

Given that you're running Windows 11 Insider (v26100) and Hyper-V is enabled, the combination of an early build, experimental driver compatibility and virtualizations likely caused the kernel component to crash.

RECOMMENDED FIXES
1. Perform a clean GPU driver reinstall using DDU
Download DDU (Display Driver Uninstaller)
Boot into Safe Mode
Run DDU to remove all traces of:
NVIDIA / AMD / Intel GPU drivers (depending on what you use)
Install the latest stable (non-beta) driver from the official GPU vendor website:
NVIDIA
AMD
Intel
Avoid Insider/Preview or Beta GPU drivers on Insider Windows builds.

2. Check for virtualization conflicts (Hyper-V, WSL2, RDP, OBS, Parsec, etc.)
Since the dump confirmed Hyper-V is active, and the crash occurred in a rendering subsystem, this could be a virtualization-induced GPU bug.

Temporarily disable Hyper-V and test stability:
Run this command in an elevated PowerShell or Command Prompt:

bcdedit /set hypervisorlaunchtype off

Reboot your PC.

To re-enable Hyper-V later:

bcdedit /set hypervisorlaunchtype auto

Disabling Hyper-V will affect apps that need it (WSL2, Android Subsystem, Docker Desktop, etc.)

3. Ensure no overclocking is applied
Overclocking GPU/CPU/VRAM can increase the chance of memory-related BSODs, especially in kernel space.

Disable overclocks via BIOS or software tools (like MSI Afterburner, Intel XTU, AMD Ryzen Master, etc.).
4. Run global system integrity checks
Run in a terminal as Administrator:

sfc /scannow
DISM /Online /Cleanup-Image /RestoreHealth

5. RAM & hardware diagnostics
Check your RAM:

Use Windows Memory Diagnostic (mdsched.exe)
or
MemTest86 (most reliable, bootable tool)
Also, monitor GPU and CPU temperatures during workload. Overheating can lead to instability.

6. If you’re using Windows Insider / Canary Builds (26100+)…
You are running a Preview build, which is inherently less stable. This dump shows the system driver dxgmms2.sys from build 10.0.26100.3912.

Options:

Update to a newer Insider build (if available)
Consider switching back to stable builds (Windows 11 23H2 – Build 22631.x) instead
WHEN does the BSOD happen?
If this crash happens during:

Gaming, high GPU usage
Idle/Desktop
Remote Desktop / OBS / screen capture / virtual machines
Then it’s likely tied to how the OS manages graphics memory under load or with software rendering interference (OBS, UWP apps, screen recorders, VMs). Disable non-essential graphics tools and monitor behavior.

I can exclude overheating in my case - I get 67 degrees in FurMark. All the other components run at good temperatures as well. The hottest part - my CPU - runs at 82-85 max whet totally loaded.

PS The strange thing is that my Win11 is NOT an insider preview...