3D Graphics, Unexpected error running DirectX 12 Test with GeForce RTX 5050

[Austin_Lien]

Hello Passmark,
I found some error message during the 3D Graphic Test and not only one unit. FR: 4/12 units.

These setting config only use CPU / Memory / 2D graphics / 3D graphics are at 50% and also open AMCAM camera app at same time.
I don't think these loads are very large, but they will easily cause problems if they are less than 4 hours.

Of course, it is also reported to NV simultaneously, but the dump generated when nv reports a fail does not see NV's error.
NV also thinks it is a burnin app problem, need passmark to help confirm.

I had attached the fail trace log, Could you help point me in the next direction? Or any suggestions?
Is it possible to open the amcap camera app while executing the burnin process?

SERIOUS: 2025-05-03 08:10:00, 3D Graphics, An error occured during the DX12 3D test
LOG NOTE: 2025-05-03 08:10:00, 3D Graphics, Unexpected error running DirectX 12 Test. Error Number 0x887A0005 (-2005270523)
LOG NOTE: 2025-05-03 08:10:00, 3D Graphics, The GPU device instance has been suspended. Use GetDeviceRemovedReason to determine the appropriate action.
LOG NOTE: 2025-05-03 08:10:00, 3D Graphics, DXGI_ERROR_DEVICE_REMOVED (0x887a0005) - Device Removed reason: DXGI_ERROR_DEVICE_REMOVED (0x887a0005)

[David (PassMark)]

The error is fairly clear.
DXGI_ERROR_DEVICE_REMOVED = The device (the video card) was removed.
Assuming you didn't physically remove the card from the machine mid test, it is likely the device driver crashed.

From the log it seems like you are running multiple video cards. Intel UHD and GeForce RTX 5050
It also seems there is load on both cards as well. So you could try disabling the each card one at a time to see what effect that has.

You might also want to have a look at the Windows Event log as driver crashes are sometimes recorded there as well.

You could also try just running the 2D and 3D tests without all the other load and without 3rd party software.

You are also using an old release of BurnInTest, version 10.2 1016
You might like to try the latest release.

[Austin_Lien]

Hi David,
Unfortunately, I changed to v11.0 build 1008 yesterday and still saw this fail.
Can we use the current FAIL TRACE LOG to clearly point out which graphics card is more related?
Because NV believes that the dump generated by force at the time of FAIL does not show any NV problem.

[David (PassMark)]

As already suggested, turn off the Intel card in BIOS.

Then there would be only one DirectX device in the system. Then if you still get an DirectX DXGI_ERROR_DEVICE_REMOVED error there is nothing else to blame except nVidia.

[David (PassMark)]

An update for this issue:

We have had a few different customer reporting issues with the new 5xxx series nVidia cards (5060Ti, 5060, 5070, etc..) starting mid April 2025.

There seems to be at least 5 separate errors these cards are getting. All seem to be real issues, but with different levels of severity and customer impact.

Some of these error codes are also documented by Microsoft in DirectX
https://learn.microsoft.com/en-us/wi...2d-error-codes

Issue 1: DXGI_ERROR_DEVICE_REMOVED (0x887a0005) - From 3D test
This is the error documented in the posts above. From the point of view of the operating system and DirectX the video card disappears or gets reset. We expect the customer impact from this to be pretty high. (e.g. screen going black while playing some games, maybe with a recovery after several seconds, as modern Windows will attempt to restart the video driver if it crashes). So in our opinion this is fairly serious, even if it recovers and doesn't crash the entire operating system.


Issue 2: D2DERR_RECREATE_TARGET, Error Number 0x8899000C (-2003238900) - From 3D test
This is actually a Direct2D error. But still a video card issue. The details from Microsoft are, "There has been a presentation error that may be recoverable. The caller needs to recreate, rerender the entire frame, and reattempt present". We suspect the common cause for this is the card or video driver crashing. But in theory it can also happen if there is some major reconfiguration of the video setup (for example a monitor resolution change, or extra monitor being added on a live system). But at best it is an interruption to the 3D tests. The software can in theory sometimes recover from this, by recreating the 3D scene from scratch, but from the user's point of view it might be a short system freeze while everything is restarted on the video card (we don't know exactly how long this freeze would be in real life, but maybe a couple of seconds. Likley depends on the load and the hardware in use). This would be kind of annoying while gaming, the user might see this as a shuddering very inconsistent frame rate. Games that handle this error correctly would likely continue to run however. DirectX doesn't provide any details about the exact trigger of these errors, so we don't know what the root cause it. Most likely cause is a video card driver bug. Or there may be a new common video reconfig trigger condition that we aren't aware of at the moment. If the error was infrequent some users may be prepared to put up with the bad behavior, or not even notice it.

Example BurnInTest log line when level 2 logging is enabled
+121.765s - ERROR: Asteroids::RenderUI mD2DDeviceContext->EndDraw HRESULT (0x8899000c)
+121.765s - DEBUG: DX12 Render call failed - code: -2003238900

These errors can also be associated with errors in the Windows event log. These event log errors come from nvlddmkm (the NVIDIA Windows Kernel Mode Driver). ID 153, with the message "Error occurred on GPUID: 100".

Example windows event log:




Issue 3: GPGPU test timeout - From GPGPU test
We've only had one report of this error with the 5xxx series cards. The GPGPU test, with runs compute tasks on the video card, fails to start when under high load. There were no actual errors from the device driver for this error. But there was no output from the test process running on the video card and after a few minutes a watch dog timer flagged it as non-responsive. Which normally means it crashed (or insanely slow). Cause unknown, but can't rule out hardware failure.

Issue 4: Verification failure of GPU buffer contents - From 2D test
Example log line:
LOG NOTE: 2025-05-27 13:17:17, 2D Graphics, GPU NVIDIA GeForce RTX 5070, Verifcation failure of GPU buffer contents (Offset: 2676915368, Read: 0x555555085555559c, Expected: 0x5555555555555555)
The value read from the video memory wasn't the same as the value written (so the data is now corrupt). How bad is this? It might result in 1 pixel on the screen having a slightly wrong colour. It might have no effect if that video RAM isn't used. It might also crash the video card driver if you get unlucky and it corrupts GPU code / shader code. If the driver or card did crash, then you might also see timeouts. If this was main system RAM, the solution is easy, you would definitely replace the RAM stick and RMA it. With the video card, they are way more expensive and it almost seems a shame to replace the entire card just for a few bad bytes, but user impact is random and totally unpredictable. Most games will probably run fine however. This could even cause some of the other issues above, if unlucky.

Issue 5: 2D Graphics - No operations reported in timeout period
We've only had one report of this error with the 5xxx series cards. The 2D test, with runs a task on the video card to check video RAM, fails to start when under high load. The checking is done via allocation a large amount of video RAM and writing data into it. Then validating the data is correct. There were no actual errors from the device driver for this error. But there was no output from the test process running on the video card and after a few minutes a watch dog timer flagged it as non-responsive. Which normally means the task running on the video card crashed (or was insanely slow). Cause unknown, but can't rule out hardware failure.


Other notes:

1. The same tests on current AMD and Intel cards don't have these issues. The 3xxx and 4xxx series nVidia cards also don't have this problem, except when there is real hardware or driver failure, which seems to be far less common that the 5xxx series cards.
2. Some of these problems are linked with high load on the video card and video memory pressure. So turning off the 2D test, which verifies video RAM works, fixes (or at least hides) some of these errors. The 2D video RAM test allocates a lot of RAM on the video card. Meaning that having low amounts of video RAM seems to contribute to these issues. But in the cases we have seen so far, there is always video RAM still available. So it isn't situation where there is no free RAM.
3. The faults doesn't seem to be linked any OEM. Some errors have been seen on Gigabyte, ASUS and MSI cards.
4. Errors occur fairly quickly if they are going to occur (for example after 15min of heavy load)
5. Swapping the suspect video card can sometimes fix the error (we would think for the VRAM error especially a swap would fix it). But for the other issues a swap doesn't always fix it. Which points to a driver issue.
6. Three different tests in BurnInTest are throwing errors (GPGPU, 2D and 3D). They all use different code. It seems unlikley all three could have newly discovered bugs simultaneously.
7. The issue seems a lot more common on the 8GB video cards (5060 and 5070) than the 16GB cards. This issue has not been reported on the 5080 nor 5090 for example. This aligns with the point above about memory pressure. It has also been widely reported that 8GB video cards in general fail on some games when running at 4K with Ray tracing, again due to low RAM. Even very low end video cards such as Intel integrated GPUs should be fine with BurnInTest however. So this doesn't explain these errors entirely. But maybe there is a driver bug for the 5xxx series card when in a low RAM situation. This would also explain some of the game crashes.
8. There is also a lot of public info on nVidia messing up with their recent video drivers. For example https://www.reddit.com/r/pcgaming/co...le_gpu_driver/