Render and compute on a cloud GPU.

Command-line rendering and batch jobs on a dedicated NVIDIA GPU. Upload your files, run the job, download the results, stop the machine.

What runs on it

Linux command-line tools that use an NVIDIA GPU through CUDA. You install them yourself; PyTorch and the CUDA libraries it uses are already there.

Blender (Cycles)

Download the Linux build and render in background mode (blender -b) on the GPU with CUDA

Python + CUDA libraries

GPU-accelerated number crunching with PyTorch, or pip install CuPy and similar libraries

FFmpeg and media tools

Install with apt-get and script conversions and batch processing

Your own scripts

Anything that runs on Linux from a terminal — you are root in the container

GamePC machines have no display, so anything that needs one won't work:

✕Programs that need a graphical desktop (e.g. After Effects, Premiere Pro, DaVinci Resolve, the Blender interface)
✕Screen sharing or streaming a desktop to your computer
✕Real-time viewport work

1Upload your project files

Use the upload button in Studio’s file browser, or scp over SSH.

2Install your tools

You are root in a Linux container: apt-get install, pip install, or download a Linux build.

3Run the job on the GPU

Start it from a Studio terminal or over SSH. Use nohup so it keeps running if you disconnect.

4Download results & stop

Download the output, then stop the machine — the charges stop. The disk is kept for 24 hours, then deleted.

The Power tier has the most GPU memory in the lineup — 48 GB — for large scenes and datasets.

Power

GPURTX A6000

VRAM48 GB

RAM64 GB

vCPUs12 cores

$1.20/hr

Charged by the second. Stop anytime.