AWS Thinkbox Discussion Forums

Concurrent Tasks and Nuke Renders

Hi there,

I am currently experimenting on how to make my Nuke renders in Deadline more efficient/faster so I have been experimenting with some Deadline settings.

Before I made any changes, I was rendering with the following Deadline settings:

-10 frames per task

  • 1 concurrent task

For a heavy render, it roughly took 10 minutes to render those 10 frames on a single machine so essentially 1 frame/minute.

I then switched the render settings expecting it would be faster on a single machine:

-1 frame per task
-4 concurrent tasks

And each frame took roughly 5 minutes to complete so I had a worse performance.

Do you have any ideas on why this might be? Or any other settings I might need to adjust?

Our machines have 128GB RAM and have AMD Ryzen 9 5900X so they should be find handling concurrent frame renders I would expect.

Thanks!

At some level it is simple - launching Nuke is an expensive operation, and batching frames together in the same run is more efficient.
The dependencies can be complex though, so this is best determined empirically.
For 10 frames I would certainly expect a single task to perform better, but it becomes different at larger volumes.

In the docs you can see all available options:
https://docs.thinkboxsoftware.com/products/deadline/10.0/1_User%20Manual/manual/app-nuke.html

There is Batch mode, which reuses the same running Nuke between tasks without relaunching it.

Privacy | Site terms | Cookie preferences