Jump to main content

Quicklinks

Technical Details

RAMSES (Research Accelerator for Modeling and Simulation with Enhanced Security) is the HPC cluster of the University of Cologne providing research groups from North Rhine-Westphalia and researchers in life sciences from Germany with a computing power of 4.8 Petaflop/s for modeling and simulation of complex problems. Especially , data from life science is securely stored and processed.

The cluster overall offers 164 compute nodes in different variants including big nodes with 3 TB main memory as well as nodes with GPUs for AI applications. Most of the nodes provide 192 cores of two AMD Genoa CPUs. Those processors are particularly well suited for high-performance scientific computing due to their high memory and interconnect bandwidths.

All compute nodes are interconnected via an InfiniBand network, which offers a bandwidth of 100 Gb/s gross at low latency. With that and the big compute nodes, RAMSES is suitable for both highly scalable MPI applications and memory consuming SMP applications.

Our documentation has moved to GitLab of HPC.NRW umgezogen. All about the RAMSES cluster and its usage you can find here:

RAMSES Documentation

RAMSES Hardware

All compute nodes are equipped with AMD EPYC CPUs of the fourth generation (Genoa) and DDR5 main memory at a speed of 4800 MT/s. The Nvidia H100 GPUs offer 94 GB HBM2 memory, the A30 GPUs 24 GB HBM2 memory. The big SMP nodes and the interactive nodes are further equipped with additional SSDs to locally store data.

  • Compute nodes
    • 136 SMP/MPI nodes with 192 cores, 768 GB RAM, 960 GB SSD
    • 8 BigSMP nodes with 192 cores, 3 TB RAM, 15.36 TB SSD
    • 8 Interactive nodes witht 192 cores, 4 A30 GPUs, 1.5 TB RAM, 15.36 TB SSD
    • 10 GPU nodes with 96 Cores, 4 H100 GPUs, 1.5 TB RAM, 960 GB SSD
  • Test nodes
    • 1 node with 128 cores, 2 MI210 GPUs, 256 GB RAM, 1.92 TB SSD
    • 1 node with 24 cores, 2 SX-Aurora TSUBASA, 192 GB RAM, 1.5 TB SSD
  • Kubernetes nodes
    • 10 nodes with 128 Cores, 1 TB RAM, 960 GB SSD
  • Storage with parallel file system
    • 15 PB HDD
    • 940 TB SSD
  • High-performance network
    • InfiniBand HDR100 with 100 Gb/s

RAMSES Software

RAMSES carries out user jobs with the SLURM batch system. You can start your jobs right away with application software already installed or you can build your own application with the development software provided.

  • BLAST+, BWA, Flye, freebayes, GATK, Hifiasm, HMMER, MrBayes, RAxML, SAMtools, STAR, Trinity, VCFtools
  • CP2K, CFOUR, Gaussian, GROMACS, LAMMPS, ORCA,  NWChem, QuantumESPRESSO, TURBOMOLE
  • Java, Julia, Mathematica, MATLAB, Miniconda3, Python, R, Rust
  • AOCL, BLIS, FFTW, GSL, MKL, OpenBLAS, FlexiBLAS, LAPACK, ScaLAPACK
  • Open MPI, Intel MPI
  • GNU, AMD, Intel, Nvidia Compiler
  • CUDA, ROCm
  • Advisor, VTune, Trace Analyzer and Collector, Forge DDT/MAP, TotalView, Vampir

Weitere Informationen: RAMSES Dokumentation