BurninTset 10.2 shows a 2D graphic test error: Error verifying data in GPU memory.

[dqa_minisforum]

Sorry, I just saw your reply, thank you.

[dqa_minisforum]

Any update? Please.

[David (PassMark)]

Log has errors like this



Many different memory locations (offsets) were affected and the memory corruption involved many bits (0x264 in binary is 0010 0110 0100, which means 4 bits are being flipped).
This isn't typical of a hardware DRAM failure. So more likely there is some software corrupting the memory.

This can happen for example when a bug results in the allocation of the same video RAM twice, to two different tasks and then they corrupt each others RAM. Typically the bug is in the video card driver. Maybe the new version of Windows is using the graphics differently, triggering the bug.

If you search through the forum you will find lots of examples of this in the past. But they are hard bugs to fix and only maybe 1 or 2 people on the planet have power to fix these video drivers. They are very hard people to get access to.

Updating the video card driver (or falling back to an older one) can also help.... sometimes.

Also trying V11 can't hurt.

[dqa_minisforum]

I recorded the debug-log when the fault occurred, please help check it, thank you.

[David (PassMark)]

Would be interesting to narrow down the problem to a combination of tests.

For example if it only happened when the DirectX12 3D test and video RAM test are run at the same time, it might be some change in DirectX that caused this.

[dqa_minisforum]

There will be no errors if only test 2D.
I have changed other brands of RAMs, and this fault also occurs, including Samsung and Hynix and so on.

[David (PassMark)]

Does the error occur if you only run the 2D test by itself?

if you collect debug logs they might contain some additional information
https://www.passmark.com/support/bitdebug.php
Logs will be smaller and cleaner if you only run the 2D test by itself.

Did you try replacing the RAM on a failing machine with new set of RAM?
Did you try a different brand of SOMIMM?

[dqa_minisforum]

The V11 has not been tested.
My hardware configuration is AMD 7945HX CPU and two 32GB SODIMMs.
The fault occurred on many machines.
This error occurs only when my machine has two 32GB RAM modules or two 48GB RAM modules,
but not when I have two 16GB RAM or only one 32GB RAM and only one 48GB RAM.
It's not random and will happen within two hours.
If the test is extended to a duration of 24 hours, it will have three or four errors.

[David (PassMark)]

You are using V10 of BurnInTest, So you might want to try V11. I wouldn't expect any change in this behavior however.

What hardware are you using?
Does this only happen on a single machine, or many machines?
If many machines are they all the same spec, or different hardware?
Does the error happen at random, or only after 1h30min of testing?
Is the error always just a single bad byte of video RAM, or sometimes you see more than one error?

[dqa_minisforum]

I have verified several times that this error occurs 100% on a 24H2OS, and it never occur when I return the OS to 23H2 , so it is not a coincidence.

[David (PassMark)]

You should
1) Reproduce the test several times and make sure this happens 100% of the time on 24H2.
2) Revert to 23H2 and check the problem never happens, even when you do several runs.

As it might just be a co-incidence.

Might just be a purely random hardware fault (or driver bug). RAM faults and software errors are not always consistent.

[dqa_minisforum]

My PC is an integrated graphics card, not a discrect graphics card, so it can't be swapped. And my PC has tested MemTest86, and the result is pass. So I don't think it's an HW issue.

[David (PassMark)]

Maybe it is a real hardware fault in that case. Did you try swapping the video card?

[dqa_minisforum]

The graphics driver was not updated and is the same version, only the operating system was updated.

[David (PassMark)]

Maybe update came with a new video card device driver version?
Did you compare the versions?