Command line rendering causes PC to freeze totally / Maya 2018.4 + VRay 3.6.0.4 + Windows 10

[seq]

Hi everyone,

we have a nasty problem with our new PCs (all AMD Threadripper 1950X & 2990WX & MSI® X399) when we batch render via batch render file (each PC renders locally - no DR): The rendering process sometimes causes the PC to freeze totally. The monitor gets black / Remote Desktop connection breaks and we can only turn off the PC with the ON/OFF button on the power supply. Sometimes this happens after 2h, sometimes after 1 day.

The PC are brand new, have enough RAM (64 GB) and are not getting to hot. We already have checked the windows system logs with our vendor and there are not hints what may cause the problem.

I have already read this topic:
https://forums.chaosgroup.com/forum/...ring-maya-2018

And tried to lower the Threads that are being used during rendering. But I'm not quite sure that VRay or mayabatch is really respect this. Can someone please tell me how I can check how many threads are used by mayabatch under Windows 10? When I look in the task mgr -> Details and enable "Threads" then it says 366?

No other software is running besides Maya 2018.4, VRay 3.6.0.4 + PhoenixFD 31.001 (without license / only for the water texture).

As suggested I will try to disable the Windows Index function.

As this happens on six PCs (all AMD Threadripper 1950X & 2990WX & MSI® X399 / Windows 10), it seems unlikely that it is hardware related.

Any other tips? This is very frustrating as the Threadrippers are so fast...

Thanks,
Lars

[seq]

Addition: The same scene renders fine on another PC also with Maya 2018.4 + VRay 3.6.04 + PhoenixFD + Windows 10 but with Intel Xeon CPU and 40 Threads. So maybe the AMD Procs are guilty?

Does anyone using a system with a Threadripper + X399 + Windows 10?

[seq]

Sorry, another addition: We have another PC also with AMD Threadripper 1950X & MSI® X399 which also renders fine. And the only difference is that it has a Geforce 1060 (because it is also used as an workstation) and the others only a GT 710 or Radeon R5. But as we rendering hardcore CPU the graphic card should not play any role, or?

[seq]

Another strange thing: On the PC that is rendering fine the CPU workload is always 95%. The other always have a workload of 100% and in the resource monitor the CPU workload is 108%? Hm, perhaps we need to reduce the workload. Anyone knows how this in done in Windows 10?

[seq]

This env variables also works for VRay for Maya, or?

set /a CORE_COUNT=%NUMBER_OF_PROCESSORS%-1
setx /M VRAY_NUM_THREADS %CORE_COUNT%

[Zdravko.Keremidchiev]

Those should work with Maya as well.

[seq]

Unfortunately they don't help at all. We will now try to clone the Windows system from the working AMD machine to one o the faulty ones. May the Windows God be with us!

[ivan.slavchev]

Is it possible that there is some overclocking set up on the motherboard or via 3rd party software? If there is - can you turn it off?

[seq]

Hi Ivan, no, no OC at all.

[seq]

Crazy thing is that we have 7 PC with AMD Threadripper + MSI X399 + Windows 10. And 1 is rendering fine for weeks and 6 freeze after irregular periods.

[seq]

All failed. Will try to update to Next and see if that helps. Perhaps the Intel road would had been the safer route.

[ivan.slavchev]

Isn't here anything in Windows Event log with an error description, or a code? Is it possible that the CPU overheats?
You can try to drastically reduce the core count - e.g. 50 % and check if that works.

You can also try running V-Ray Benchmark - it loads the CPU heavily - and see if it'll crash again - https://www.chaosgroup.com/vray/benchmark

Do you get the same error if you render from inside Maya (instead of batch)? Or - if you try using only V-Ray - DR for example, or V-Ray standalone with a .vrscene file.
You can also try disabling simultaneous multithreading for the processor and see if you'll get the same issue.

[seq]

Hi Ivan,

thanks for your suggestion.

Fortunately - knock on wood! - it seems that after a installation marathon the problem went away.

I don't know which step really solves the problem, but that is what we have done:

- Update Windows 10 to the current state of Version 1803

- Update to VRay Next Hotfix 1

- limited the threads with this code:

set /a CORE_COUNT=%NUMBER_OF_PROCESSORS%-2

setx /M VRAY_NUM_THREADS %CORE_COUNT%


To answer your question:

- there was no error description at all in the Event logs

- no CPU overheating

My feeling says it was a software problem. Either Windows related or VRay/HT related. But of course that is only a guess. But as the hardware is not faulty it must have to do with the software.

Man, we are so happy that this hopefully is solved!

By the way we just rendering an aerial view of a city with thousands of lights and VRay does a great job! Thanks Chaosgroup!

Lars

[ivan.slavchev]

Hi Lars,

Thanks for the feedback, great that you were able to fix it. It sounded like a Software problem since I haven't heard of a particular problem with AMD Threadripper but just wanted to rule out some possibilities. An option in such cases is to enable Windows Crash dumps and hope to see something there, but no needed anymore, fortunately.

[waaazoo]

Another strange thing: On the PC that is rendering fine the CPU workload is always 95%. The other always have a workload of 100% and in the resource monitor the CPU workload is 108%? Hm, perhaps we need to reduce the workload. Anyone knows how this in done in Windows 10?

Hi Lars. I had a similar 'some kind of) problem recently, also with threadripper. I had no computer freeze. but noticed that my CPU was never at 100%. It ended up, after some testing, that it was a heat problem. My Watercooling was defect and the CPU was at 68°C and throttled even down. Since I've put back the Noctua Aircooling my PCU is rendering at 100% at 50 to 55°C. Threadripper throttles down automatically at 68°.
I know it's not really the problem you had, but if it ever comes back, then you could also check your temps (which you obviously did, reading your last post)