GPU rendering crashing machine

[johnBKQP5]

After using Arnold gpu on one project over the winter.. which while idiosyncratic, was pretty great, however on this next project can't get it to work at all. It often crashes my machine.. if not Maya.. or it simply stops working ('render aborted' with no errors). The render usually only uses 5-6gb vram (have 32gb). CPU works flawlessly.

I've tried DDU'ing, reinstalling/rolling back graphics drivers, reinstalling MtoA, clearing optix caches.. messing with TDR registries. I've tried without my NVlink bridge/SLI enabled.

Pretty much nothing I do gets gpu rendering working again unless I open a simple plane, sphere and skydome env lgt file (I'd once made as a 'reset') and flush all caches.. switch to cpu.. render it, save the file, then restart the machine.. reopen the file, flush caches, prepopulate, then switch to gpu.. but even that simple file will eventually get corrupted and stop working after a few tries. This routine was the idiosyncratic issue with my first maya/arnold gpu project, btw, but I did at some point get into a reliable groove with it and finished that project. This project has a heavier asset (texture-wise) but I still have never cracked 10gb vram.

It now seems the TDR regedit may be preventing the hard crashes, but Maya will still crash or abort the gpu render every time. (Btw, Substance Painter can render with iRay GPU all day long on the same asset, same hdr with no problems).

Latest Maya2020, latest MtoA, latest Nvidia RTX driver (..as of 3/15/21). ASUS Pro WS x570-ace Ryzen 3950x with 128gb (non ecc) ram and 2x Nvidia RTX5000 w NVLink/SLI, 1300w PSU, latest Win10 Pro.

****

* Arnold 6.2.0.1 [903992ac] windows clang-10.0.1 oiio-2.2.1 osl-1.11.6 vdb-7.1.1 clm-2.0.0.235 rlm-12.4.2 optix-6.6.0 2021/02/09 10:40:42

* CRASHED in RaiseException at 00:00:00

* signal caught: error E06D7363

*

* backtrace:

>> 0 0x00007fffc1b3d75a [KERNELBASE ] RaiseException

* 1 0x00007fff9cb46220 [VCRUNTIME140] CxxThrowException

* 2 0x00007fff56225ff5 [ai ] Ordinal0

* 3 0x00007fff568c4eed [ai ]

* 4 0x00007fff568ca0c1 [ai ]

* 5 0x00007fff56a8d84d [ai ]

* 6 0x00007fffc1e81bb2 [ucrtbase ] configthreadlocale

* 7 0x00007fffc35a7034 [KERNEL32 ] BaseThreadInitThunk

* 8 0x00007fffc4262651 [ntdll ] RtlUserThreadStart

*

* loaded modules:

* 0x00007fffc1b10000 KERNELBASE

* 0x00007fff9cb40000 VCRUNTIME140

* 0x00007fff56180000 ai

* 0x00007fffc1e60000 ucrtbase

* 0x00007fffc3590000 KERNEL32

* 0x00007fffc4210000 ntdll

****

Stack trace:

ntdll.dll!RtlRaiseException

KERNELBASE.dll!RaiseException

VCRUNTIME140.dll!_CxxThrowException

ai.dll!Ordinal0

ai.dll!AiUnknownFunction2

ai.dll!AiUnknownFunction2

ai.dll!AiUnknownFunction1

ucrtbase.dll!_configthreadlocale

KERNEL32.DLL!BaseThreadInitThunk

ntdll.dll!RtlUserThreadStart

00007FFF8EA0F600Result: C:/Users/johnh/Documents/maya/projects/c130/scenes/c_130_debug_FILE__5060.ma

Fatal Error. Attempting to save in C:/Users/johnh/AppData/Local/Temp/johnh.20210314.2004.ma

// Maya exited with status 1

[adrien.herubel]

Hi John,

Regarding the crash, did you submit a crash report ? If so, we should be able to find it and investigate it directly.

Could you set the log verbosity levels in the Arnold render to "Debug", and share the resulting log on a failed GPU render ? Alternatively if you can't see the error in Maya, can you export an ass file and use kick

If the crash scene specific, we can also try to repro using the scene if you are able to share it with us.

[johnBKQP5]

ok I will try that and submit in the next few days once i've got a break, thx