Cycles GPU backend cost optimization on Blender 4.2 LTS

Goal

Determine the most cost‑effective Cycles backend for sporadic rendering jobs on a low‑traffic pipeline that runs on Blender 4.2 LTS.

Constraints

The pipeline uses a cloud instance that can be spun up on demand, but idle GPU time between renders incurs significant electricity and rental charges. The scene complexity is modest, and the render duration is typically under five minutes. The available backends are CUDA, OptiX (for NVIDIA RTX), and CPU.

Uncertainties

Blender 4.2 LTS may have shifted backend performance, and the cost of GPU versus CPU hosts depends on the specific cloud provider’s pricing model. There is no built‑in scheduler to shut down instances automatically.

Questions

• What is the comparative wall‑clock time for a 5‑minute render on a single‑core CPU versus a CUDA or OptiX GPU on a 4.2 LTS build, and how does that translate into hourly cost?
• Does Blender 4.2 LTS support automatic backend selection based on scene statistics, or must the backend be set manually in Preferences?
• For a headless render on a spot‑instance, does the cost of idle GPU time outweigh the performance benefit over a CPU‑only instance?