Dedicated render servers as slave nodes for Deadline 8, maximizing limited resources

Deadline
1

Hello,

We currently have 6 slave nodes in the studio, and are using our workstations (mostly higher consumer range quality components) as slaves, usually after working hours, to render files from Maya with VRay renderer, either with distributed rendering, stand-alone rendering, or via job submission on Deadline 8 farm.

Jobs are very diverse, can be from single very high quality, high resolution renders to high quantity HD rendering.
We are currently rendering from a 1300 render layer file, taking cca 20 hours to render on our best workstation.

How can we get the most out of our workstations?
We use CPUs, Xeons and i7s, OS is Windows 10. Can we use GPUs (consumer grade Nvidia cards) to do the rendering using Deadline? If so, are there any limitations compared to CPU rendering?
How can we speed up the pre-render phase (opening scenes, calculating stuff) on Deadline?

Would it make sense (time and money wise) to build a dedicated render slave machine(s)?
What would be best focus for such a build? Go for max CPU power or maybe for multi-GPU solutions?
What would be a sweet-spot, best price performance point for such a build?
Would it make sense to use Windows OS or maybe consider some linux distro?

Thank you for reading so far, and for your replies and interest.

Klemen

2

Hey Klemin,

This post would likely be more at home on studiosysadmins.com/ I believe than here as this is a multi-faceted set of questions. I can give some advice on what I know from the Deadline workflow front, but as I have very little real-world experience with V-Ray I can only tackle it from the Thinkbox tool perspective. In fact, most of your questions would focus on V-Ray since it’s going to be doing most of the heavy lifting.

The one thing I can advise here is that when rendering still frames, tile rendering is going to net you some serious benefits versus just letting it cook on a single machine. You did mention pre-calculating here, so you may already be doing this (V-Ray has some gamma issues without it).

For OS option, I believe it makes more sense to go with what you’re comfortable with. Often, saving time debugging unrelated problems when thing go wrong is worth the cost of the Windows license for example. Same goes for if you’re a Linux guy.

As far as creating dedicated render nodes, it may be worthwhile if you need immediate feedback on renders, but you can use lower-res renders during the day and feed the farm the big stuff overnight.