OctaneRender community
Multi-GPU OctaneBench configurations
Multi-GPU configurations published in community OctaneBench results, together with custom chassis manufacturing and long-term support.
GRUENCY / Rendering workstations
OctaneRender workstations for GPU rendering
GPU rendering speed grows with the number of GPUs, provided PCIe lanes, cooling and noise are planned for them. GRUENCY plans all three before any part is ordered. Sebastian Michalski has built multi-GPU OctaneRender workstations since 2011 and is known in the OctaneRender community as smicha.
What is decided before ordering
- GPU count sized to scene memory and target frame times
- PCIe topology planned slot by slot across four to seven or more GPUs
- Noise target for a working studio, agreed at specification
- Custom water-cooling loop designed for sustained render load
- Storage layout for assets, caches and output
- Remote commissioning into your render pipeline
Testing under render load
Each rendering workstation goes through a sustained full-load burn-in, a thermal profile for every GPU and a noise check against the studio target. It is delivered with its own test record.
How an OctaneRender workstation is sized
- VRAM per card comes first. Every GPU holds the whole scene, so the largest scene sets the memory per card.
- Card count comes second. Octane scales across GPUs close to linearly, so frame time falls with each card running at full clocks.
- OctaneBench is the throughput target, measured on the finished workstation under sustained load.
- Power and cooling are sized for all cards running together for hours. Water-cooling keeps clocks and noise stable.
- The same workstation also runs Redshift, Cycles and other GPU path tracers.
OctaneRender and OctaneBench history
Prototype, 2017
11 water-cooled GPUs in one chassis
A dual-CPU server platform in a single case, built in 2017.
NVIDIA vendor work
Visual production systems
Vendor work for NVIDIA on VFX, rendering and visualisation workflows.
Questions about rendering workstations
Which render engines do you build for?
GPU path tracers in general. Most of the work has been with OctaneRender, and the same configurations run Redshift and similar pipelines.
How loud will it be in a studio?
A noise target is agreed at specification. Custom water-cooling keeps a multi-GPU workstation quiet enough for a studio.
Can you build several nodes?
Yes. Small render farms and rack systems go through the same testing and are commissioned remotely into your pipeline.
How much GPU memory does an OctaneRender scene need?
Octane renders each frame on every GPU in parallel, so the scene has to fit in the VRAM of each card. VRAM does not add up across cards. Out-of-core textures and geometry help, at a cost in speed. GPU choice starts from the largest scene you render. The card count then follows from the frame times you need.
What does OctaneBench tell you about a workstation?
OctaneBench measures path-tracing throughput. Octane scales across GPUs close to linearly, so the OctaneBench score of a multi-GPU workstation is a practical guide to how many frames it renders per hour. This holds when PCIe lanes, power and cooling keep the cards at full clocks under sustained load.
Who is smicha?
smicha is Sebastian Michalski, founder of GRUENCY and of the WS.COMPUTER hardware line. He is known under that name in the OctaneRender community for multi-GPU, water-cooled render builds.
Request a consultation
Describe the scenes, the render engine and where the machine will be used.
Or email info@gruency.com