What GPU generations are available on Saturn Cloud?
Saturn Cloud provides access to H100, H200, B200, and B300 GPU instances. H100 and H200 are available across multiple regions via AWS, GCP, Azure, Nebius, and Crusoe, while B200 and B300 Blackwell instances are available via Nebius.
How does Saturn Cloud handle data security?
Saturn Cloud deploys inside your own cloud account, ensuring your data never touches their servers. It offers VPC deployment, full network isolation, SOC 2 compliance, encrypted data, audit logging, and private networking with no public endpoints.
Can I use my own Docker images and ML frameworks?
Yes, Saturn Cloud supports custom Docker images, allowing you to bring your own dependencies and frameworks. It runs standard Python, so frameworks like PyTorch, HuggingFace, and vLLM work as-is, with CUDA and drivers pre-configured.
How does Saturn Cloud compare to AWS SageMaker?
Saturn Cloud runs standard Python without proprietary APIs, unlike SageMaker's SDK. It supports multiple clouds (AWS, GCP, Azure, etc.) while SageMaker is AWS-only. Saturn Cloud also offers H100s from $2.95/hr, often more affordably than SageMaker's premium pricing.
Does Saturn Cloud support multi-node GPU training?
Yes, Saturn Cloud supports multi-node clusters for distributed training workloads, including FSDP, DDP, and DeepSpeed. You can provision these clusters from the dashboard without manual node configuration, utilizing NVLink 4.0 for inter-GPU communication.