Dear Users of SuperMUC-NG Phase 1!

In August, SuperMUC-NG Phase 1 has completed its 7th year of operation. Installation and the first Top500 entry was even a year earlier. Its hardware has aged and is now out of support. We are aiming to keep the system alive and in operation until the end of the year 2026, however, at slightly reduced capacity. The accelerated SuperMUC-NG Phase 2 system is expected to remain in operation until mid of 2028. 

We kindly ask all active projects to consider the following:

Allocated Compute Time: Your granted CPU hours can only be consumed until the end of operation of SuperMUC-NG Phase 1. They will not be "reimbursed" nor taken over to another machine. They cannot be consumed on SuperMUC-NG Phase 2. No further CPU time will be granted except for small allocations to finalise a project. Be aware that the machine is heavily overbooked for the remaining time of operation, so try to consume your budget as soon as possible.

Future CPUh Allocations: New proposals for CPU-only projects can be submitted at TIER 2 level to NHR. Here, access is restricted to researchers from German universities. Scientists from other research institutions can apply at the compute centres of their organisations. In addition, all researchers can check for applications at the European level with EuroHPC.

Port to Accelerators/GPUs: SuperMUC-NG Phase 2 contains Intel Ponte Vecchio GPUs and is targeted to operate at least until mid 2028. The NVIDIA Vera-Rubin based Blue Lion system is planned to be operated at least for five years, once it becomes available. For both architectures a variety of community and commercial software has already been ported. If you plan to port your custom code to GPUs we offer a number of relevant courses. Additional support regarding code profiling and porting is offered by the Computational X Support group at LRZ (open a ticket or reach out to one of the application labs). Financial support for porting activities is offered by the KONWIHR initiative limited to Bavarian researchers.

Please note: CPU-only applications will not be accepted on accelerated systems. A reasonable initial GPU usage is expected.

Secure Your Data: Along with the compute nodes also the file systems WORK and SCRATCH have aged. The risk of failures has increased and repair times may be longer as spare parts are not readily available. While we will keep WORK under hardware support for the remaining operation time of SuperMUC-NGPhase 2, SCRATCH is only supported  by best effort and total data loss is an imminent danger. We advise all projects to regularly back-up their current simulation campaigns and archive data of finished projects. With the Data Science Archive we offer plenty of archive capacity for SuperMUC-NG users. 

If you have any questions or require consulting concerning your HPC project, please reach out to us via the servicedesk.lrz.de or visit us in the weekly HPC Lounge. 

Frequently asked Questions

Will my remaining compute budget be transferred to SuperMUC-NG Phase 2 or the upcoming Blue Lion system?

No.

SuperMUC-NG Phase 2 has a very different architecture compared to Phase 1. And the upcoming Blue Lion system will be very different again. Transfer/exchange of compute budget is therefore not possible.

Phase 2 gets almost all its compute power from the Intel Ponte Vecchio GPUs (Intel Data Center Max 1550). Applications therefore must be able to use the GPUs.

Can I continue my project on SuperMUC-NG Phase 2?

Please submit a Ticket via the Service-Desk to activate access to Phase 2.

All projects on SuperMUC-NG Phase 1 receive a compute budget of 900 GPU-h on the Phase 2 system (the same amount as if you applied for a test project). You can use this to test if your application runs on Phase 2. 

I need to finish important research - can you prioritize my jobs?

Unfortunately, no.

All jobs on SuperMUC-NG (Phase 1 and Phase 2) are considered equally important. Prioritizing one over another is not possible.