Sie befinden sich im Service: RWTH High Performance Computing (Linux)

Introduction to CLAIX-2025

Introduction to CLAIX-2025

information
 

Temporary page with ongoing changes!

This page is designed to aid users during the pilot phase of the new Claix-2025 cluster by navigating you through the common pitfalls.

CLAIX-2025 is the latest addition to our cluster environment and features modern hardware to run your HPC and machine learning workloads. As part of our 1-Cluster concept we are making sure that working with our different clusters simultaneously works as easily as possible.


Common questions

How do I connect to CLAIX-2025?

CLAIX-2025 has its own login nodes that you can use to prepare and post-process jobs. The login nodes share the same software environment and hardware topology as the batch nodes in the HPC and ML segments:

Who can submit jobs to CLAIX-2025?

Unlike CLAIX-2023, job submission is restricted to Tier-2 NHR and WestAI projects. Have a look at our project catalogue to check whether you are eligible for a project application. Access to the login nodes is not restricted to specific projects. Jobs can be submitted from all login nodes using one of the CLAIX-2025 partitions in your job script. Jobs submitted from the new login nodes will be submitted to CLAIX-2025 by default. 

How do I share data between both clusters?

CLAIX-2023 and CLAIX-2025 share their entire personal and project storage so you can access data and work concurrently on both clusters. 

Do I need to recompile my own software for CLAIX-2025?

Not necessarily. The CPUs used in CLAIX-2023 and CLAIX-2025 are mostly compatible with each other. Depending on the build configuration and how aggressive your optimization settings were, software you compiled for CLAIX-2023 may not run (performantly) on CLAIX-2025. In these cases, we recommend using a different build for the new cluster. 

I am experiencing problems with my jobs on CLAIX-2025

As with every new cluster, we may need to adjust system parameters and fix problems depending on user feedback. We have run extensive functionality and performance tests on the cluster but cannot cover the shear breadth of applications running on our systems. If you experience problems on CLAIX-2025 that cannot be fixed by recompiling your software, please inform us via servicedesk@itc.rwth-aachen.de. Your feedback is valuable!

Changes in the module system

CLAIX-2025 uses its own software tree optimized for AMD CPUs and the OmniPath network system connecting the different nodes of the cluster. As a result, there are a few divergences in the list of available software on CLAIX-2023 and CLAIX-2025 due to incompatible software versions or deprecated modules that were not ported to CLAIX-2025 to begin with. 
Besides these differences, we are using the pilot phase to introduce several mostly non-intrusive enhancements to our module system, some of which will soon be introduced to CLAIX-2023 as well:

Lmod 9

We have upgraded Lmod for CLAIX-2025 to the latest major version which comes with a variety of bugfixes and also a new feature that suggests how to proceed if you want to load a module that requires a different compiler or MPI runtime. 

Compiler modules are now mutually exclusive

When we migrated to Lmod on CLAIX, we explicitly did not make compiler modules mutually exclusive to maintain some backward compatibility to workflows that were common on our old system. This has caused a lot of confusion, especially for newer users, and is thus corrected for CLAIX-2025. This means that trying to load two compiler modules such as GCC and intel-compilers will cause the first module to be unloaded alongside with all dependent modules that are not available for the new compiler. This leads not only to cleaner module environments but prevents you from loading potentially incompatible module combinations. Combining other compilers with GCC is still supported, GCC is just not set as the default compiler.

zuletzt geändert am 13.07.2026

Wie hat Ihnen dieser Inhalt geholfen?

Creative Commons Lizenzvertrag
Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Deutschland Lizenz