Cluster GPU rendering

[infoLG6NH]

Hi everyone,

I'm having trouble setting up Cluster GPU rendering. We're using version 2025.1. One workstation is running VRED with 2 GPUs, and we want to use a second workstation (also with 2 GPUs). We have "activate cluster service" enabled, and the VRED Cluster Setup seems to be configured correctly. Licenses are fine (2x Pro), and there are no error messages. The cluster can be set up and activated (both nodes are yellow).

When we start rendering, there's network activity on the remote workstation (so the project is being transferred to the node), and both GPUs' VRAM on that remote station fills up. However, the GPU utilization is either zero or very sporadic. In fact, rendering ends up being slower than when using a single workstation.

We last used this setup with version 2024 (not sure which exact 2024.x), and everything worked perfectly.

Does anyone have any idea what might be causing the issue? Any help would be greatly appreciated.

Thanks in advance!
Michal

[michael_nikelsky]

What is the network and hardware configuration? First thing I would always do is set the image compression setting in the cluster to Medium or High, no matter how fast your network is, it will be the bottleneck. 

The other thing is: What are your scene settings? Some things like Glow/Glare/DLSS/Denoising require buffers to be transfered as 16bit floating point and DLSS or Denoising also require more than one buffer to be transfered. The PCIE and the network will cause a massive performance limit in these cases, so much that especially with DLSS and Denoising I would advice to not use a cluster for interactive rendering since in my experiences it was always slower unless you are rendering really heavy scenes which only run at less than 10fps anyways (Offline rendering is fine since the buffers only need to be transferred after the last image sample was rendered).