Case Study: Potsdam Institute for Climate Impact Research achieves faster, reproducible climate modeling on HPC with Anaconda

How Anaconda Helps PIK Teams Steward a Brighter Future

The Potsdam Institute for Climate Impact Research (PIK) is a non-profit climate research center with over 400 staff that runs a large HPC cluster (≈5,000 CPU cores across 350 nodes, ~200 teraflops). Facing the need for flexible, reproducible, high-performance tooling across disciplines (data science, physics, oceanography, economics, etc.), about 100 PIK users have relied on the Anaconda platform (including conda and Intel MKL integration) since 2015 to manage Python packages and environments.

Anaconda enabled PIK to decentralize software provisioning and deliver optimized numerical builds, so scientists can create environments in minutes instead of waiting hours or days for admin support, and admins spend less time building dependencies from scratch. The result: faster model turnaround, smoother machine‑learning workflows (e.g., installing latest PyTorch/CUDA for GPU training), improved reproducibility, and broad adoption across ~100 users on PIK’s HPC cluster — all achieved with Anaconda.

Potsdam Institute for Climate Impact Research

Ciaron Linstead

Software Support Engineer