High Performance Computing (HPC)¶

What this service is¶

UCT High Performance Computing (HPC) provides shared compute infrastructure for research workloads that are too large, too slow, or too resource-intensive to run effectively on a laptop or standard workstation.

Use HPC when your work involves:

large datasets
long-running analyses
parallel or distributed computation
GPU-based workloads
batch workflows that need to run reliably on shared infrastructure

When to use HPC¶

HPC is a good fit when you need to:

run analyses that take hours or days
process data at a scale that exceeds local machines
request CPUs, memory, GPUs, or wall time explicitly
separate interactive work from production-scale runs
move from exploratory work to repeatable batch execution

If your work is lightweight, short-running, or mostly document-based, another service may be more appropriate.

If you are unsure where to begin, see Choose the right service.

What HPC provides¶

HPC gives researchers access to:

shared CPU and GPU compute resources
scheduled batch execution on compute nodes
interactive and browser-based access options
high-performance storage integration
a platform for running research software at scale

UCT HPC uses scheduled job submission rather than direct execution on shared infrastructure, and the reference section already separates core topics such as scheduling, storage, GPU access, graphical applications, and Open OnDemand.

How HPC fits with other services¶

HPC is usually part of a broader workflow rather than a standalone service.

It is commonly used together with:

Research data storage for active project data and HPC-linked storage
Data transfer for moving data between systems and collaborators
Research software for managing code, scripts, and workflows

Start from your task¶

If you already know what you are trying to do, go to:

Understand how HPC works¶

Go to Reference when you need concepts, system behaviour, or constraints:

Need help?¶

If you need support, go to: