DXVK 2.6.1/3.0.x black screen in Assassin's Creed Origins and loading crash in Watch Dogs 2 on Intel Iris Xe
I remember when I tested a few years ago, at least DXVK 2.3 was working and really smooth.
I have the exactly same processor, let me download the game back to test again.
BTW, driver version above 7000 makes my device randomly frozen, so I'm using 6737 from Microsoft driver pack.
Yes I reproduced crash on my end. It always crashes near loading end
I'll do a bisect.
Release Bisect:
2.4.0 works
2.6.0 works
2.6.2 works
2.7.0 stuck at loading, sometimes crash after stuck
Thus, it should be some regression issue introduced between 2.6.2 and 2.7.0
EXCEPTION_DEBUG_INFO:
dwFirstChance: 1
ExceptionCode: C0000005 (EXCEPTION_ACCESS_VIOLATION)
ExceptionFlags: 00000000
ExceptionAddress: disrupt_64.00007FFC95388E5D
NumberParameters: 2
ExceptionInformation[00]: 0000000000000001 Write
ExceptionInformation[01]: 0000000000000002 Inaccessible Address
First chance exception on 00007FFC95388E5D (C0000005, EXCEPTION_ACCESS_VIOLATION)!
15 call stack frames (RIP = 00007FFC95388E5D , RSP = 000000003235EA40 , RBP = 000000003235EB50 ):
000000003235EAC8 disrupt_64.RunGameEx+3EA0D
000000003235EBA8 disrupt_64.RunGame+3688
000000003235F408 disrupt_64.enum AKRESULT __cdecl AK::Monitor::PostCode(enum AK::Monitor::ErrorCode, enum AK::Monitor::ErrorLevel)+5C9E711
000000003235FCC8 disrupt_64.WaitForActivation+998CBE
000000003235FCF8 disrupt_64.enum AKRESULT __cdecl AK::Monitor::PostCode(enum AK::Monitor::ErrorCode, enum AK::Monitor::ErrorLevel)+71F9A6C
000000003235FD28 disrupt_64.WaitForActivation+7A7212
000000003235FD68 disrupt_64.enum AKRESULT __cdecl AK::Monitor::PostCode(enum AK::Monitor::ErrorCode, enum AK::Monitor::ErrorLevel)+5D59444
000000003235FDB8 disrupt_64.enum AKRESULT __cdecl AK::Monitor::PostCode(enum AK::Monitor::ErrorCode, enum AK::Monitor::ErrorLevel)+5DD2F2C
000000003235FE68 disrupt_64.WaitForActivation+12146D6
000000003235FE98 disrupt_64.void __cdecl AK::MemoryMgr::GetPoolMemoryUsed(long, struct AK::MemoryMgr::PoolMemInfo &)+216E4D
000000003235FEF8 disrupt_64.enum AKRESULT __cdecl AK::MemoryMgr::SetMonitoring(long, bool)+BFDD2D
000000003235FF28 msvcr110._beginthreadex+107
000000003235FF58 msvcr110._endthreadex+192
000000003235FF88 kernel32.BaseThreadInitThunk+17
000000003235FFD8 ntdll.RtlUserThreadStart+2C
Continue bisect
02c6caabcfeb5fb1b8cf0084bae7d21755514049: crash
05d69eb689e6b05da7a3cc62d99c0fd318a6350b: works
b90c526df02f21b053a693e27f1010383c14f3d2: crash
23d9cb2bc294fc4870cd6e22e5473092beb95a8f: crash
bbe3a692d24f95a3cf1f89f24a228fb7b7507c20: works
bc488c042932478e28e5a118cf4e24a0416497fa: works
e7b909c03fc6f30c1d15cfd514128026c9196c94: crash
So, I can confirm commit e7b909c03fc6f30c1d15cfd514128026c9196c94 introduced this regression issue.
Thank you so much for doing a full git bisect to isolate issue. I honestly didn't expect anyone to go through all that trouble to test this on an i5-1135G7!
I am currently on driver 7085, which is almost certainly why I am getting these issues.
I am going to cleanly roll back my Intel driver to the pre 7000 branch (i found a 32.0.101.6987 version) and test DXVK 2.6.2. Really appreciate the time and effort you put into helping me troubleshoot this!
Btw did 3.0 DXVK work on your specs? Thanks again.
Sounds like a whole slew of driver bugs, none of these things are an issue elsewhere (including Intel ANV).
Don't think we can really do anything here (short of rewriting the binding model yet again, which is obviously not really an option).
Btw did 3.0 DXVK work on your specs? Thanks again.
No, same crash.
BTW, if you want a somehow old but the most stable driver version I have ever tried, use this https://www.microsoft.com/en-us/download/details.aspx?id=103503
Microsoft chosen driver version did work best, while after upgrading to any 7000+ version my Surface frequently frozen when rotating screen.
Sounds like a whole slew of driver bugs
I agree, it's just yet another Intel driver quality issue and they have made so many such issues.
@Anuradha-117
Maybe we should report to https://github.com/IGCIT/Intel-GPU-Community-Issue-Tracker-IGCIT , but considering Intel has ceased support 11-14th, don't think will be fixed.
Greetings, @Anuradha-117
I am DXVK-GPLALL developer. Please, can you check DXVK-GPLALL 2.6.8-2? I am interested in your feedback.
Greetings, @Anuradha-117
I am DXVK-GPLALL developer. Please, can you check
DXVK-GPLALL 2.6.8-2? I am interested in your feedback.
it works on my end
@qinlili23333
Thanks for the report.
@qinlili23333 Thank u for your support. Btw I found 2.6.2 doesn't have this issue, and as you said it works (But FPS are lower by about 5 than dxvk 1.10.3). But anyhow I now have Linux mint installed on a different partition and use that one. No more issues.
@Digger1955 I will check that out and report back. DXVK 2.6.2 worked as @qinlili23333 said so this might work too.
Nothing extracted yet.
Software information
Assassin's Creed Origins
Watch Dogs 2
System information
DXVK versions tested
Assassin's Creed Origins
Watch Dogs 2
Native D3D11 works in both games.
Symptoms
Assassin's Creed Origins
With DXVK 1.10.3, the game loads and runs normally, but memory usage appears to increase over time. After some time (usually during the 3rd benchmark run or after 10-25 minutes of gameplay), the game exits to desktop without an error message.
DXVK 2.6.1, 3.0, and 3.0.2 do not reach gameplay. After pressing "Continue", the loading screen finishes but the game remains on a black screen indefinitely.
Watch Dogs 2
DXVK 1.10.3 works correctly. DXVK 3.0 and 3.0.2 crash near the end of the loading screen.
Logs
Logs from both games are attached.
ACOrigins_d3d11.log
ACOrigins_dxgi.log
WatchDogs2_d3d11.log
WatchDogs2_dxgi.log
Steps to reproduce
Assassin's Creed Origins
Watch Dogs 2
Additional notes
I am unsure whether this is a DXVK issue or an Intel Vulkan driver issue, but I wanted to report it since DXVK 1.10.3 works while newer versions do not.