What is High-Performance Computing (HPC)?

Overview

• Time: 15 min

  1. Understand what High-Performance Computing (HPC) is.

  2. Learn when to use HPC.

  3. Explore real-life examples of HPC applications.

High-Performance Computing (HPC) refers to the use of supercomputers and parallel processing techniques to solve complex computational problems at high speeds. HPC systems are designed to perform large-scale computations that require significant processing power, memory, and storage capabilities. They are used in various fields such as scientific research, engineering simulations, financial modeling, and data analysis.

A laptop or desktop computer typically has limited processing power and memory compared to HPC systems. For example, a high-end laptop might have

• 16 GB of RAM

• 8-core CPU

• GPU with 4 GB of memory

In contrast, an HPC system is much more powerful. for instance, Gadi, the supercomputer at the National Computational Infrastructure (NCI) in Australia, has:

• 48 cores per node

• 192 GB of RAM per node

• Four NVIDIA V100 GPUs per node each with 32 GB of memory

Explanation

Node refers to a single computing unit within an HPC system. Each node can have multiple CPUs and GPUs and independent memory.

A cluster is a group of interconnected nodes that work together as a single system. Each node typically has its own CPU, memory, storage, and possibly GPU resources. The nodes are connected through a high-speed network, and they coordinate to run tasks in parallel.

An HPC system is a very powerful cluster.

Moreover, there are many nodes in the system, which can work together to perform computations in parallel. This allows HPC systems to handle large datasets and perform complex calculations much faster than a typical laptop or desktop computer.

When to Use HPC?

HPC is used when the computational requirements of a task exceed the capabilities of standard computing systems.

• When you are dealing with large datasets that does not fit into the memory of a single machine.

• When your application/program is computationally intensive and requires significant processing power.

• When your application is time-sensitive and requires fast execution.

Real-life Examples of HPC

HPC is used in various fields to solve complex problems, simulate real-world scenarios, and analyze large datasets. Some exxamples where Gadi, the supercomputer at NCI Australia, is used include:

Bush Fire Simulation

The Bureau of Meteorology, using Gadi, is creating advanced modelling tools that are already enhancing authorities’ understanding of fire dynamics. Fire agencies now use these models in real time to forecast how fires might spread.

Drug Discovery

Drug development companies and universities are simulating protein structures, human cell receptors, and more, with a focus on designing new drugs and vaccines. Other research groups, backed by universities, are also contributing to this effort.

Genomic Research

Genetic medicine is set to benefit from innovations enabled by NCI and the Gadi supercomputer. The Garvan Institute is updating its 4,000-genome Medical Genome Reference Bank, a key comparison set of healthy genomes. In 2017, NCI processed 1,000 genomes overnight—a major milestone.

Computational Modeling

Capturing flame behavior in engine conditions with full detail is nearly impossible without computational modelling. The most accurate method—Direct Numerical Simulation (DNS)—uses full physical and chemical equations to simulate each step, requiring millions of compute hours. Researchers at UNSW run DNS on the Gadi supercomputer. Their recent work involved grids with over 1 billion points, 10,000 time steps, and 50+ variables, using up to 20,000 processors per simulation.

Satellite Imagery Analysis

ANU and Geoscience Australia reanalysed 30 years of Landsat images to create the Barest Earth map—showing Australia with minimal vegetation. Built from petabyte-scale data stored at NCI, the map reveals bare soil across the continent and provides a key dataset for future research.

Weather Simulation

The Bureau of Meteorology ran a high-resolution weather model over Australia using a 400-metre grid—spanning 5,600 km east-west and 4,000 km north-south—with 12.6 billion points. Using the Unified Model on NCI’s Gadi supercomputer, the simulation took 3 weeks, 12,000 processors, and generated nearly 100 TB of data—setting a new scale record for the model.

Airflow Simulation

University of Melbourne is using the Gadi supercomputer to simulate airflow in turbines and generators with high precision. The goal is to improve future designs for greater efficiency, lower emissions, and longer lifespans.

Climate Modeling

NCI’s data environments support large-scale climate analysis using scalable HPC methods and shareable Jupyter notebooks. Enhanced search and collaboration tools let researchers efficiently explore CMIP6—the Coupled Model Intercomparison Project phase 6, an international collaboration producing standardized climate model simulations.

Key Points

• HPC systems are designed to perform large-scale computations that require significant processing power, memory, and storage capabilities.

• HPC is used when the computational requirements of a task exceed the capabilities of standard computing systems.

• HPC is used in various fields such as scientific research, engineering simulations, financial modeling, and data analysis.