To support our diverse compute-intensive workloads, we operate a modern distributed-memory cluster infrastructure based on Dell PowerEdge systems, combining Intel- and AMD-based compute nodes within a unified high-performance environment.

The compute nodes in this cluster span multiple hardware generations and are optimized for both general-purpose and memory-intensive scientific applications. The newer nodes (installed in spring 2025) are based on the Dell PowerEdge R7625 platform and are equipped with dual AMD EPYC 9355 processors, providing 64 physical cores (128 threads) per node. These nodes feature high-speed DDR5 memory, with configurations ranging from 512 GB up to 1 TB per node, enabling efficient execution of highly parallel and memory-demanding workloads. Selected nodes are additionally equipped with NVIDIA L40S GPUs, providing modern GPU acceleration for compute-intensive tasks such as machine learning and data analytics.

Complementing these are compute nodes based on the Dell PowerEdge R740 platform with dual Intel Xeon Gold 6248 processors. These nodes provide 40 physical cores (80 threads) per node and are available in both standard-memory configurations (192 GB RAM) and high-memory configurations (768 GB RAM), supporting a broad spectrum of simulation and data-processing workloads. A subset of these nodes is equipped with NVIDIA Tesla T4 GPUs, which provide additional acceleration capabilities for specialized workflows.
Across all Slurm-managed compute nodes, the cluster provides a total of 992 physical CPU cores (1984 logical CPUs), offering substantial computational capacity for scientific computing. The system is connected via a high-performance interconnect and provides centralized RAID-based storage for temporary simulation data, forming a flexible and scalable HPC environment for simulation, modeling, ray tracing, data analysis, and AI-driven applications.

AMD-based compute nodes (6x)
PowerEdge R7625 Server
2x AMD EPYC 9355, 32C/64T per CPU
64 cores / 128 threads per node
512 GB DDR5 memory

AMD-based high-memory / GPU nodes (2x)
PowerEdge R7625 Server
2x AMD EPYC 9355, 32C/64T per CPU
64 cores / 128 threads per node
1 TB DDR5 memory
NVIDIA L40S GPU

Intel-based compute nodes (10x)
PowerEdge R740 Server
2x Intel Xeon Gold 6248, 20C/40T per CPU
40 cores / 80 threads per node
192 GB DDR4 memory
480 GB SSD
Intel Omni-Path Host Fabric Adapter
Selected nodes equipped with NVIDIA Tesla T4 GPUs

Intel-based high-memory nodes (2x)
PowerEdge R740 Server
2x Intel Xeon Gold 6248, 20C/40T per CPU
40 cores / 80 threads per node
768 GB DDR4 memory
480 GB SSD
Intel Omni-Path Host Fabric Adapter

Login/storage node
PowerEdge R740XD Server
2x Intel Xeon Gold 5217 (8C/16T)
128 GB RAM
24X 2 TB SAS
 PERC H730P+ RAID Controller
Intel Omni-Path + 10 Gbit Ethernet

Network infrastructure
100 Gbit/s Intel Omni-Path interconnect
10 Gbit/s Ethernet management network
Dell S4148T-ON and H1024-OPF switches