Choose your GPU.
Compare memory, hardware, and hourly rates. Choose a GPU that fits your workload.
Explore the hardwareCompare GPUs for training, inference, and rendering. Choose your hardware and configure your workspace in the console.
NVIDIA & AMD.
Your own code.
Tools you already use.
Compare GPU memory and current hourly rates. Filter by workload to find a starting point for your setup.
Checking GPU availability…
Availability observed over the last 60 minutes. Capacity can change; review the current region, GPU count, and rate before launching.
Choose PyTorch for training, Jupyter for notebooks, or Ubuntu for your own environment. Preview a template before configuring your instance.
Configure a workspaceChoose a template. Bring your code.
Pick the tools your project needs.
import torch device = torch.device("cuda")model = YourModel().to(device) # Train on the GPU.model.train()$ python train.py --device cuda
↳ CUDA device ready.
Compare memory, hardware, and hourly rates. Choose a GPU that fits your workload.
Explore the hardwareChoose your region, GPU quantity, and template. Review the included disk, name your instance, and select an SSH key.
See the workspaceConnect to your instance, track its cost, and terminate it when you finish. Your workspace keeps the activity history.
Open your consoleWhat to know before you start.
Explore PomPod
Compare the hardware, choose a template, and try your setup in the console.
Find your GPU