Using multiple GPUs for rendering on AWS

We are doing rendering with “movie render queue” using an AWS server

EC2 types: from “g4dn.8xlarge” up to “g4dn.metal”

GPUs: Tesla T4 (have tried also v100 with p4 instances)

AMIs: “NVIDIA RTX Virtual Workstation - WinServer 2019” - Marketplace , basic Windows server 2019 with 466.47_vGPUSW_vgaming_win10_server2016_server2019_64bit_international driver (installed according to AWS docs)

tool: movie render queue

We try to speed up the rendering process, can anyone answer the following questions?

  • max 50-60% of GPU is used while rendering, why?

  • is there a way to use more than 1 GPU while rendering?

    • I have come across with Nvidia SLI, but that feature seems disabled on AWS according to “GPU-z” program with both marketplace and regular AMIs, how can I enable it?

    • Is there anything else/ other way?

Or maybe you could suggest another way to speed up the process?

I was interested in this as well. Did you find a solution?

Still would love some insight into using multiple GPUs with MRQ. Is it possible? Which rendering modules support it if so?