National supercomputer Roihu

Scientific high-performance computing

A supercomputer is an important research tool in modern science and the use of supercomputers, or high-performance computing (HPC), is becoming more widespread in all fields of science. Having our own national supercomputer ensures that all researchers have access to competitive HPC resources and significantly speeds up research. Supercomputers are also used in university teaching.

Roihu will replace CSC’s current national supercomputers Mahti and Puhti and it will be in researchers’ use in March 2026. Roihu will triple the current performance of Finland’s national supercomputing resources.

Roihu’s pilot testing will begin in February 2026, and it will be generally available in March 2026. The call for Roihu pilot projects is open until November 28.

Roihu is a highly versatile platform for researchers across various fields:

• Most researchers using CSC’s AI-related services train small to medium-sized models, for which Roihu is well suited.
• We are developing the capability in Roihu to process sensitive or confidential data.
• Researchers can use it for example to analyze audio and video materials, compute molecular dynamics simulations, screen potential drug compounds from extensive libraries, simulate glaciers and calculate climate scenarios.
• The educational use of HPC in higher education institutions is a new and growing field. While it doesn’t demand extensive computing resources, it involves a large number of users. This approach helps cultivate the next generation of researchers across all disciplines and prepares skilled experts for industry.

What is a supercomputer?

A supercomputer is one of the world’s fastest-performing computers. One of the criteria is making it to the list of the world’s top 500 supercomputers. The computing power of supercomputers is expressed as flops, or calculations per second. This abbreviation stands for floating point operations per second. A floating point number is the number format used in computing.

The performance of a supercomputer depends largely on its high-speed interconnect network and memory capacity rather than exclusively on its power and number of processors. Two types of processors are used in supercomputers: CPUs and GPUs. Of these, central processing units (CPU) are conventional general-purpose processors.

A newer and increasingly popular GPU (graphics processing unit) was originally developed as a graphics processor. Today, GPUs are widely used in supercomputers, particularly for processing large datasets and artificial intelligence. Software in other scientific fields has also advanced significantly in recent years, allowing GPUs to be utilized more broadly than ever before.

The latest development focus is on combining high-performance computing, quantum computing and artificial intelligence.

The life cycle of a supercomputer is approximately six years.