Crashes

[coilbook]

Hi,

Thank you for the feedback. I am looking through the materials you provided and seeking to duplicate the issue.

Thank you

I think it happens more often if you have lots of trees, grass, scattering medium, lights and render times over 3-4 min per frame.

[Goldisart]

maybe it is banal overheats ?????

[coilbook]

maybe it is banal overheats ?????

I guess it could be. Our cards run at 65-75C
The only problem this morning netstor box that hold 4 780 TI cards was still using 1 cards while other 3 failed. We never had this in version 2. Usually if one fails all cards stop and rendering stops. But once we hit cancel render 3ds max just crashed and closed as always and this never happened in ver 2.

[Goldisart]

could still be the problem of motherboard I had to change the Asus for the Gigabyte and it worked .... this was a lot of problems with 3.- 1. Blue screen of death 2. The computer freezes 3. Stop the render sequences.

...

p.s. the temperature is 75 degrees Celsius I think it's normal

[coilbook]

I am not sure if it is mother board. All 3 slaves give us errors 719 randomly and they worked fine until we switch to version 3. Also cancel render and max crashes never happened before in version 2. Hopefully future updates fix it.

[coilbook]

as soon as I read your message I got this. before it would freeze up like it is almost finished rendering a frame, When i closed the window max just closed. We never had this before
We use 3ds max 2015 windows 7, 362 drivers. maybe scattering medium in the sun is doing it. They promised to fix this bright sun problem when mixed video cards are used in AO mode. viewtopic.php?f=9&t=55380&p=283639&hili ... rs#p283639

[coilbook]

One more thing
We use netstor box with 4 GPUs and we've been having where 1-2 GPUs would stop rendering and the other 2 continue. Never had this problem in version 2

can it be because of this:
"Since the beginning of Octane the integration kernels had one CUDA thread calculate one complete sample. We changed this for various reasons, the main one being the fact that the integration kernels got really huge and impossible to optimize. Also OSL and OpenCL are pretty much impossible to implement this way. To solve the problem, we split the big task of calculating a sample into smaller steps which are then processed one by one by the CUDA threads. I.e. there are a lot more kernel calls are happening than in the past."

parallel samples value is too high maybe some cards get stuck calculating longer and fail.

[azen]

One more thing
We use netstor box with 4 GPUs and we've been having where 1-2 GPUs would stop rendering and the other 2 continue. Never had this problem in version 2

can it be because of this:
"Since the beginning of Octane the integration kernels had one CUDA thread calculate one complete sample. We changed this for various reasons, the main one being the fact that the integration kernels got really huge and impossible to optimize. Also OSL and OpenCL are pretty much impossible to implement this way. To solve the problem, we split the big task of calculating a sample into smaller steps which are then processed one by one by the CUDA threads. I.e. there are a lot more kernel calls are happening than in the past."

parallel samples value is too high maybe some cards get stuck calculating longer and fail.

Hey,

One potential issue you could be having is with some stability issues with the latest CUDA drivers. Try lowering the CUDA drivers on your master and slave to 358.50, and let me know how that goes.

Before you do that, here are some logging files that will help to get more valuable information. Download the attached files and copy octane_log_flags file into the standalone install directory on the master and slave machine. Also copy the octane_daemon_log_flags file into the install directory of the slave machine as well. This will output information into a file named "octane_log". So far, it is not easy to reliably duplicate your issue based on the info you provided. The more information we can use, the better. Also, let me know if reverting your CUDA drivers to 358.50 resolves your issue.


Also