Re: DXGI_ERROR_DEVICE_REMOVED or DXGI_ERROR_DEVICE_HUNG

So right now, there is a very large thread on this forum related to the DXGI crashes happening with Frostbite. 

It's really difficult to debug for regular users as it happens randomly, and most commonly in REDSEC (at least for me) meaning I am risking ruining the games of my friends by constantly joining and crashing every game. Without even so much as a rejoin button playing with me becomes more of a burden than anything else. 

This issue got a lot worse after Season 2 started and despite the huge amount of feedback from the community on this issue, there appears to be very little response from EA, DICE, Nvidia and AMD on the issue. 

At the very least an acknowledgement about the issue and some general pointers as to what is more likely to cause it allowing users to take some mitigating steps would be enormously helpful, but instead we get radio silence. 

DXGI_ERROR_DEVICE_REMOVED or DXGI_ERROR_DEVICE_HUNG | EA Forums - 12737991

There is the thread in question. 

PROCESS_NAME:  bf6.exe

ERROR_CODE: (NTSTATUS) 0x887a0006 - The GPU will not respond to more commands, most likely because of an invalid command passed by the calling application.

EXCEPTION_CODE_STR:  887a0006

STACK_TEXT:  
00000000`7fbfa310 00000001`44d96671     : 00000000`00000044 00000000`3229b180 00000000`326d0820 00000000`7fbfa615 : KERNELBASE!RaiseException+0x8a
00000000`7fbfa410 00000001`44d96d04     : 00000000`6cc355b0 00000000`06f3e9f8 00004990`1d66e54b 00000001`4531c765 : bf6!ffxFsr2ResourceIsNull+0x22c5431
00000000`7fc0f610 00000001`44d954e3     : 00000000`6cc355b0 460f5dcf`36555f00 00000000`7bf1d2f0 c3c3bfa7`1bb45900 : bf6!ffxFsr2ResourceIsNull+0x22c5ac4
00000000`7fc0f660 00000001`44c3dba6     : 00010001`40000020 3a6a99de`20ce91f2 00000000`6cc32cf0 00000000`06f3e9f8 : bf6!ffxFsr2ResourceIsNull+0x22c42a3
00000000`7fc0f770 00000001`44c3aeb8     : 00000000`6cc32cf0 00000000`6cc32cf0 00000000`06f3e9f8 00000000`6cc32cf0 : bf6!ffxFsr2ResourceIsNull+0x216c966
00000000`7fc0f920 00000001`43c06821     : 00000000`7fc0fa00 00000001`dfd33900 00000000`00000000 00000001`dfd339bc : bf6!ffxFsr2ResourceIsNull+0x2169c78
00000000`7fc0f9c0 00000001`43c0e091     : 00000001`49f77a98 00000001`43c0cbc1 00000000`00000000 00000000`7fc0fbe0 : bf6!ffxFsr2ResourceIsNull+0x11355e1
00000000`7fc0fab0 00000001`43c0d700     : 499062a6`11ab0000 00000000`7fc0fd80 00000001`49d1ed80 ffffffff`ffffffff : bf6!ffxFsr2ResourceIsNull+0x113ce51
00000000`7fc0fb30 00000001`43c0d269     : 00000000`4c6e5c44 00000000`00c60100 00000000`00000000 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0x113c4c0
00000000`7fc0fc50 00000001`43c0b863     : 00000000`00040000 00000000`00001000 00000000`00000000 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0x113c029
00000000`7fc0fcb0 00000001`43c0ac57     : 00000000`00000000 00000000`7fc0ff00 00000000`32751800 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0x113a623
00000000`7fc0fe20 00000001`436fc4ab     : 00000000`00000000 00000000`00000000 00000001`49d052d0 00007fff`be75a9ab : bf6!ffxFsr2ResourceIsNull+0x1139a17
00000000`7fc0fe70 00000001`436fe274     : 00000000`32751800 00000000`327517c0 00000000`00000000 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0xc2b26b
00000000`7fc0fec0 00000001`4330d68c     : 00000001`49a0c800 00000000`00000000 ffffffff`00000103 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0xc2d034
00000000`7fc0ff00 00007fff`bf7e37b0     : 00000000`4c5c98b0 00000000`00000000 00000000`00000000 00000000`00000000 : bf6!ffxFsr2ResourceIsNull+0x83c44c
00000000`7fc0ff30 00007fff`c11ee8d7     : 00000000`00000000 00000000`00000000 00000000`00000000 00000000`00000000 : ucrtbase!thread_start<unsigned int (__cdecl*)(void *),1>+0x30
00000000`7fc0ff60 00007fff`c1e2c3fc     : 00000000`00000000 00000000`00000000 000004f0`fffffb30 000004d0`fffffb30 : kernel32!BaseThreadInitThunk+0x17
00000000`7fc0ff90 00000000`00000000     : 00000000`00000000 00000000`00000000 00000000`00000000 00000000`00000000 : ntdll!RtlUserThreadStart+0x2c

Clearly there is one or more FSR calls happening despite FSR being disabled. This is not reflected in the DXGI error dialog for me by the way, only when analyzing the dump file. 

This issue seems to be with BF6, the AMD drivers (in my case) or both in that FSR seems to be asked to handle one or more resources that don't exist. Given that Nvidia users also have reported crashes in the thread linked above, I am inclined to believe this is a bug in Frostbite (assuming their dump has a similar stack trace)

The question is what causes these null pointers? That's where we are in the dark, and perhaps the devs could provide some insight. Is there any way to completely force the FSR2 pipeline not to be involved in the rendering at all? Could it be resources the rendering pipeline needs are unavailable due to external tools like overlays, hooks, Javelin etc?

Just... any feedback. Anything at all. An acknowledgement of the issue. Heck, I'd settle for a "Calm down, love."