Metro Exodus - c0000005 - memory access violation crash after 20 hour?

[SLitw]

Monitor:AORUS AD27QD

Cpu: i7-9900K stock 4.7 with Noctua NH-D15

Gpu: Rtx 2080 Ti Aorus Xtreme Waterforce ( stock no oc,temps max 63C on load )

Ram: 32gb 2x16 GB DDR4 GSKILL 3000MHZ XMP

Psu: Seasonic Ultra Prime Titanium 850W

mobo: Asus Prime Z390-A

SSD: Crucial 1TBHDD: WB BLUE 1 TB

Case: Cooler Master Cosmos C700P

Hello. I have a question. I left running Metro Exodus for hours ,when i go to work. When i back after 21 hours i see it crashed exactly on 20 hour in running. I left running game in EPISODE:DESERT ( 3 ).

Crash is something about c0000005 - memory access violation.

NULL POINTER WRITE.

Not driver crash. Event log nothing about driver crash. Just BUGTRAP crash.

Screen from debugger:

Is my hardware ok? I dont have any crashes in other games or benches. Realbench passing too ,memtest no errors. I am on newest nvidia drivers.Using MSI AFTERBURNER + RTSS.

[n_scott_pearson]

I think your hardware is just fine. I believe that this is a software bug.

While there are other possibilities, the most likely scenario is that a memory allocation operation failed -- but the invoking code did not check to see if a failure occurred -- so the NULL pointer returned in these cases was then used as the destination for a memory copy operation. I have seen this time and again in poorly-written C code. I have also seen it in C++ code where the writer simply assumes that an automatic memory allocation operation cannot fail (for example, an object is declared on the stack of a subroutine and the instance of the object is supposed to be automatically created (allocated) when this subroutine is invoked) when, in fact, this is *always* a possibility.

Next, let's talk about where this error is occurring. It looks like a worker thread within Metro Exodus has made a call down into the NVIDIA PhysX libraries to a function named NvHairWorks_Create() and the error occurred when this function invoked Standard C Library function memcpy(). Of course, this function is just where the error was detected - where the bad (NULL) pointer was used as destination for the memory copy operation. It is unclear whether the bad memory allocation operation occurred within this function or within the code that invoked this function. Again, this is just my take on the most likely scenario. Other possibility is a software bug that somehow loads a NULL pointer into the variable passed to memcpy().

To be honest, I am not surprised that this kind of error is occurring after 20 hours of operation. If it follows my scenario, this software probably has a small memory leak somewhere in regularly-executed code and it takes this long for the amount of memory leaked in total to build to the point where a memory allocation could fail. Someone didn't test their code for long enough!

Your task now is to communicate with the developers of Metro Exodus and have them look into the error. If it turns out that the bug is in the NVIDIA PhysX code (which I rather doubt but it is always a possibility), it would be the responsibility of Metro Exodus to inform NVIDIA.

...S

[SLitw]

OK happened again ( second time ) but this time after 12 hour, the same crash. Crashdumps are in Metro Exodus save games folder.

But i tested memtest86,RAMTEST,REALBENCH and no issues.

Only in Metro Exodus this happen on chapter 3 DESERT. No in other games.Other games i can run 24h no crash.

So no worries?

[n_scott_pearson]

Well, this points to the issue being within the Metro Exodus game itself (as opposed to the underlying NVIDIA PhysX libraries), so no worries from the standpoint of your hardware. You do need to contact the Metro Exodus folks to let them know about this issue. Who knows, they may already have a fix for this issue and can provide you with an updated version to try...

...S

[SLitw]

Hello. Thank you for reply. Somebody from other forum said to me this:"I'm a pro developer. Nothing should crash or OOM if left for 24 hours."

So i think he is not right?

[n_scott_pearson]

I have 40 years of experience, so I outrank him.​ 😋

[SLitw]

OK happened again but this time after 30 hours. Crashdumps are in Metro Exodus save games folder.

Crash is something about c0000005 - memory access violation.

NULL POINTER WRITE.

Not driver crash. Event log nothing about driver crash. Just BUGTRAP crash.

So propably game bug?

But i tested memtest86,RAMTEST,REALBENCH and no issues.

No in other games.Other games i can run 24h no crash.

[n_scott_pearson]

Yes, it's a bug in the game. Send these dumps to the Metro Exodus team for analysis.

...S

[SLitw]

Somebody said this to me :"any game should not be crash no matter how long you play or idle..."

And he tell me that is hardware.

So he is right or not?:)

But like i said,it happens only in Metro Exodus after 12-30 hours it will sometimes crash with this.

[AndrewG_Intel]

Hello SLitw

Thank you for posting on the Intel® communities.

We found that you have another thread where you are describing issues with Metro Exodus Game*. The original thread link is the following:

https://forums.intel.com/s/question/0D50P00004YP2g1

In this case, we kindly recommend you to continue the support through the original thread in order to avoid duplication of effort.

Regards,

Andrew G.

Intel Customer Support Technician

A Contingent Worker at Intel

[n_scott_pearson]

I don't believe it's a hardware problem. This guy's comment has nothing to do with the reality of the situation - and I question his "experience" if he is making a blanket comment like this. If this was the case, you would be seeing problems while other programs were running. When it is consistently isolated to a single program like this, the hairy finger points at this program until it is proven to be something else. As I said, send the logs to the Metro Exodus developers and have them to do an analysis. They can look at the source code and trace back to where the NULL pointers came from.

...S

[SLitw]

Last question. Somebody said that this is because my memory is on XMP profile. Trust him or just trust to you mr scott pearson :)?

[SLitw]

Somebody said that this is caused by XMP profile enabled. So its the cause or just game code buggy?

[n_scott_pearson]

Well, he is not right - but nor is he completely wrong. If you are running on an XMP profile, you may be running the memory buses at higher frequencies than the processor was designed and validated to maintain. If this is the case, you are technically overclocking the processor and, as we know, Intel does not guarantee operation at these frequencies. Seeing a failure like you are, which is only occurring in Metro Exodus, is simply too specific to be the result of memory failures caused by higher-than-sustainable memory bus frequencies. You would be seeing other errors, system lockups and even spontaneous system reboots in this case. Bottom line, while part of what he is saying is rooted in reality, I believe the rest is not.

Look, the proof is in the pudding. Go into BIOS Setup, turn off support for XMP profiles and then run your game test again. It will either fix the issue (which I doubt) or it won't.

...S

[SLitw]

Hey Mr scott pearson.

Yeah i had one spontaneous reboot after crash in Metro Exodus after 20 hours. Its my topic:

https://forums.intel.com/s/question/0D50P00004YP2g1/weird-crash-and-after-moment-system-restarted-on-9900k-and-2080-ti

Can you answer ? Is my ram faulty?

[n_scott_pearson]

If you are also seeing spontaneous reboots (which you have not mentioned before), then yes, ​you likely have bad DRAM (or a failure in your processor, but very unlikely).

...S

[SLitw]

No it happened once.

After 21 hours, it crashed when i press ESC and then clicked EXIT TO MAIN MENU--->moment when GPU LOAD drops it crashed ,instead of EXIT TO MAIN MENU ofc.

I get BLACK SCREEN and no monitor signal for 1 minute, and crash to desktop with TDR driver crash in event log. After that i log off from system and log back in again and system rebooted unexpected.

But log from metro crash dumps was the same ,before reboot:

The same log:

Crash was about c0000005 - memory access violation.

NULL POINTER WRITE.

Folder was created: Windows/livekernelreports/pow32kwatchdog.dmp 

If pc rebooted unexpectedly after crash no reason to worry about hardware? After crash i make CTRL ALT DELETE ,log out from system and log back in again, then pc crashed /rebooted. It happened once after 21 hours when i tried to exit to main menu. Thx

Please answer to that two questions. Is hardware fine?

No have errors in Memtest86 or RAMTEST and no issues in benchmarks and games.

[n_scott_pearson]

Again, if you are also seeing spontaneous reboots (which you have not mentioned before), then yes, ​you likely have bad DRAM (or a failure in your processor, but very unlikely).

...S

[SLitw]

But i cant reproduce that reboot again. Also not have issues with other games and no errors in memtest86

Found that file :

windows/livekernelreports/ pow32kwatchdog.dmp.

A callout to Win32k did not return promptly"

[n_scott_pearson]

Look, we are going in circles here and I only have so much time available to help other folks. MemTest86 is not infallible; it does not catch all issues. Your best bet is to find someplace to borrow some compatible memory and test with it. If it does not exhibit the issue, then formally purchase replacement memory. Your local PC repair shop may be able to help you with this process for a small fee.

...S

[SLitw]

Can you tell me what is that file. ? But i found this in windows/livekernelreports/PoW32kwatchdog.dmp <======== ( 1,8GB )

"a callout to win32k did not return promptly" windbg read

I think it was related to that reboot.