Running Notch blocks with Affinity multi-GPU

General TouchDesigner Discussion
#1

Hi team,

I’m working on a project which is at a stage where the system architecture requires some clarification on how Notch Blocks inside of TouchDesigner behave in a mutli-gpu environment.

The plan is to have two instances of TouchDesigner, each running a Notch block. Each instance of TouchDesigner is running on a separate Quadro GPU using Affinity.

GPU 1 - TouchDesigner 1 > Notch Block 1
GPU 2 - TouchDesigner 2 > Notch Block 2

I was hoping that someone might be able to clarify whether the Notch block would adhere to the affinity command and render on the specified GPU.

Thanks!

#2

Hey Jason,
Yes, it should in general. We do our best to open the same DirectX device as matches our OpenGL device. This is the same as what occurs for Spout and other things that interop with DirectX such as Nvidia Flow.

#3

Hi Jason,

I’ve validated this approach on a machine with 3 Notch blocks in 3 TouchDesigner instances across 3 RTX 6000 gpus, successfully running scenes at 21,190x1080 total canvas resolution.

Feel free to DM me if you have any more specific questions, as I designed/programmed/managed an infrastructure based on this for over a year.

Good luck!

1 Like
#4

Ah this is brilliant to hear! I’m running a similarly large resolution, so it’s nice to know this workflow is feasible.
Did you use anything like SycnCHOPs to sync the different instances of Notch, or did you let your blocks do their own thing independently?