Page 1 of 1

Cuda Error 999 failed to run post prozessing

Posted: Wed Mar 16, 2016 1:23 am
by miohn
Hi,

one Slave Computer with 2 x 980Ti (Slave Daemon Win 10 with newest nvidia drivers)
gives me that error:
Cuda Error 999 on device 1 an unknown internal error occured
Kernal syncronisation failed
device 1: failed to run post-processing

Last night it renders fine and now I suddenly get this error.
What does this mean?

thanks
Mike

Re: Cuda Error 999 failed to run post prozessing

Posted: Wed Mar 16, 2016 9:47 am
by miohn
maybe because of the newest nvidia ....364.51?

Over night only "one" frame was rendered.
Only 2 samples were missing but Octane never finished the second frame.

I installed then 362.00 which worked fine on the first Slave last weeks, but same problem occures.
After restarting the daemon - only one GPU (from 2 GPUs) is available.

???

regards
Mike

Re: Cuda Error 999 failed to run post prozessing

Posted: Wed Mar 16, 2016 3:55 pm
by bepeg4d
Hi Mike,
which plugin version?
i'm guessing, if you are on v3 alpha 6, could you try with or without the new camera imager parameter "Gamma before response"?
ciao beppe

Re: Cuda Error 999 failed to run post prozessing

Posted: Wed Mar 16, 2016 7:36 pm
by miohn
Hi,

no, all version (plugin and daemon slaves) are 2.25 latest stable.

Problem is, that the whole render prozess stops when one slave fails.
Therefore I render now with the 2. slave with only one Gpu out of two.

Maybe my PSU (700 W) from the newly added slave PC is not strong enough for 2 GTX 980 TI.

Atm I have no other idear or guess.


regards
Mike

Re: Cuda Error 999 failed to run post prozessing

Posted: Thu Mar 17, 2016 3:34 pm
by bepeg4d
Hi Mike,
700W could be not enough... have you tried to render on the slave as a master alone? If the issue persist, then it's not related to net render but, more probably, to the specific machine :roll:
ciao beppe

Re: Cuda Error 999 failed to run post prozessing

Posted: Thu Mar 17, 2016 5:02 pm
by miohn
Hi beppe,

no. There is no C4D or plugin installed on this slave.

When I first start the daemon and netrendering it is o.k. for are
few minutes until this message comes:
Cuda Error 999 on device 1 an unknown internal error occured
Kernal syncronisation failed
device 1: failed to run post-processing

When I then restart the daemon - only one GPU is provided.
But with that 1 GPU slaves is rendering since 2 days and night now.

But not with both GPUs.

What does this error "post processing" mean?

thanks
Mike

Re: Cuda Error 999 failed to run post prozessing

Posted: Thu Mar 17, 2016 5:27 pm
by bepeg4d
Hi Mike,
extra power is needed to apply the post effect, probably you are hitting the limit :roll:
what I was meaning was to render via Standalone on the slave for some time to understend if the issue is PSU dependent or not ;)
You can save the scene as Animated package (.otbx) from your master c4d and transfer it to the slave to be render in Standalone.
if you use a site like this:
http://outervision.com/power-supply-calculator
you should be able to better understand if your PSU is enough or not :roll:
ciao beppe

Re: Cuda Error 999 failed to run post prozessing

Posted: Thu Mar 17, 2016 6:38 pm
by miohn
Hi beppe,


"extra power is needed to apply the post effect,"

this I wanted to hear and confirmed the suspicion of too less power from the PSU.
But I was not sure, because before I have had exactely this new nvidia driver
installed for what Ahmet was warning.

Will look for a 1000W. This should be enough for 2 GTX980Ti.

thanks
Mike