Page 1 of 1
2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 4:55 am
by coilbook
here is error I got
I noticed after stopping rendering, going back to octane viewport, then rendering like 5 frames going back ... but after everything was adjusted and final rendering started i got this message after 3rd frame
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 5:02 am
by abstrax
coilbook wrote:here is error I got
I noticed after stopping rendering, going back to octane viewport, then rendering like 5 frames going back ... but after everything was adjusted and final rendering started i got this message after 3rd frame
The first CUDA error is the problem: For some reason the direct lighting kernel failed. If you can reproduce this problem, please send us the scene so we can investigate. Thanks.
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 5:07 am
by coilbook
thank you i will
i noticed scenes that load to memory fast crash a lot more often than the ones that takes awhile i guess cards get too hot.
Before i had a scene where it took 30-50 seconds just to load the scene before rendering. So really card had 30-50 seconds break time to cool off and it never had cuda errors .
This scene loads within seconds and i guess cards get too hot even 65C I guess is too much for GPU even though it is supposed to work up to 90C
is it possible to create like a 'break time schedule' where I can tell rednerer to stop after 30 minutes of rendering for 2 minutes to cool cards off and resume rendering. Would be nice for overnight renderings. or if it stops with cuda error Auto restart itself after a minute.
thank you
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 5:57 am
by face_off
Have you tried backing off the clock speeds of your cards a little?
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 2:31 pm
by coilbook
face_off wrote:Have you tried backing off the clock speeds of your cards a little?
i just did i think it helped
Seems like it became more stable and render time stayed the same I thought it would be slower
Thank you
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 10:19 pm
by coilbook
some lost samples on slave in 2.15
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 10:37 pm
by face_off
Seems like it became more stable and render time stayed the same I thought it would be slowe
I had a card which did a similar thing. I think it's a hardware fault - when the clock speed is too high the render starts too quickly and "something" on the card is not up to speed so it fails. Different versions of Octane have fractionally different start times, so some are effected by the problem more than others. I'm sure there is a more technical explanation for this. An Octane configurable "delay" might help with this.
Paul
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 10:59 pm
by coilbook
face_off wrote:Seems like it became more stable and render time stayed the same I thought it would be slowe
I had a card which did a similar thing. I think it's a hardware fault - when the clock speed is too high the render starts too quickly and "something" on the card is not up to speed so it fails. Different versions of Octane have fractionally different start times, so some are effected by the problem more than others. I'm sure there is a more technical explanation for this. An Octane configurable "delay" might help with this.
Paul
you might be right
because it started after installed second card 980 (I have one is 780 TI and one is 980). And it runs at a lot higher speed than 780 TI
Re: 2.15 failed to deallocate device memory
Posted: Tue Dec 16, 2014 11:00 pm
by coilbook
if it help all cards are classified evga with acx cooling and bigger board