"Orion acted instead as a drop-in platform for remote workstations that shifted our focus from finding stability to using the stability to iterate."
Donald Strubler
Head of Technology, R3D Studios, 2025
The Company
R3D Studios renders the kind of visual effects that audiences see on the screen of a feature film and assume are simply real. Their pipelines combine real-time previz, GPU-accelerated rendering, and physical simulation across teams collaborating in multiple time zones.
With 10 artists running AI-accelerated 3D conversion workflows on AWS around the clock, they delivered feature-quality work in weeks. When they came to Orion, their goal was to double headcount without adding a single GPU node.
That goal was a hard production requirement, and it’s what brought them to Orion.
The Challenge
Before Orion, R3D's infrastructure forced a tradeoff between stability and velocity.
Workstation provisioning took up to 60 minutes. Every artist start was a delay before any creative work could begin. GPU utilization was inconsistent across artist shifts: instances sat warm and billing while work paused between sessions. When something broke, the engineering team stopped shipping and started firefighting.
The deeper constraint was structural. Traditional architecture required one dedicated GPU instance per artist, regardless of actual utilization patterns. As headcount increased, infrastructure costs scaled in lockstep, and every new team member triggered a manual provisioning cycle that could take 24 to 48 hours.
Launching a workstation wasn't self-service. It was a ticket.
For a studio trying to stay lean and move fast, that was a ceiling on growth.
The Solution
R3D ran Orion as their unified compute plane, orchestrating GPU workloads and rendering jobs across the same fabric, directly in their own AWS environment. Their data stayed in their own infrastructure.
GPU time-slicing on existing instances doubled effective capacity without new hardware. Ten artists ran concurrently on five GPU nodes. The same pool that served Orion Workspace sessions during the day routed to the render queue overnight, so capacity stopped sitting idle.
Workstations launched in 60 to 90 seconds. Orion Workspace gave artists pre-configured environments they could start themselves, with no Kubernetes knowledge and no IT ticket. Admins defined templates once, and artists launched from them.
New team members onboarded themselves, and the infrastructure team got their week back.
The Results
R3D measured the results over six months of production on Orion, April through September 2025.
GPU density reached up to 2:1, with ten concurrent artists on five g5.2xlarge instances. Compute costs fell by up to ~40%, measured against standard on-demand instance pricing with no Savings Plans, Reserved Instances, or promotional credits applied. Workstations launched in 60 to 90 seconds.
Across that run, the platform had zero critical failures and no infrastructure tickets across time zones.
At 2:1 density, R3D could add artists without adding GPU nodes, which was the goal that brought them to Orion.
"Being able to treat our own internal cluster like a mini-cloud and platform for users is a game changer."
Donald Strubler
Head of Technology, R3D Studios, 2025
AT A GLANCE
FIELD
VFX & Animation
DEPLOYMENT
Customer-hosted
ABOUT
R3D Studios delivers high quality feature-film stereoscopic conversion. Their global team supports GPU-intensive rendering, simulation, and compositing across teams collaborating around the clock from multiple time zones.
