TO TOP

Simulation cluster


Introduction

Molecule
© RESOLV, Heyden

All groups using the simulation cluster facility share a common interest and quest to understand "molecular systems within the framework of quantum mechanics".

Fundamental to this endeavour is the synergistic combination of experiments that address properties site-specifically at the single molecule level with advanced computational approaches. 


Applications and Examples

Cluster-coll-3-512x384
© ZEMOS

The theoretical groups cover altogether a wide spectrum in theoretical chemistry, which reaches into the realms of theoretical biology and theoretical physics. Specific research areas are dynamical ab initio simulations of complex many-body systems at finite temperatures, the development of response methods, photochemistry and optical properties, biomolecular simulations on large time- and length-scales and the development of neuronal network methods for the heterogenous catalysis.

Enlarge image

Set-Up

Cluster-coll-2-512x384
© ZEMOS

The simulation clusters are housed in a dedicated state-of-the-art machine room with massive water cooling capabilities. For obvious reasons the electric power supply is installed under the ceiling whereas the water cooling installation is located within the double-floor below the cluster racks. The dissipated heat is also used to heat the building.

Enlarge image

 Simulation Cluster 1:

  • Launch: 2024
  • Total CPU Cores: 6,272
  • 26 CPU Nodes
            2 × AMD EPYC 9654 (2.4 GHz, Boost up to 3.7 GHz)
            2 × 96 cores (192 cores per node)
  • 10 GPU Nodes
            2 × AMD EPYC 9334 (2.7 GHz, Boost up to 3.9 GHz)
            2 × 64 cores (128 cores per node)
            4 × NVIDIA Ada L40 GPUs (48 GB RAM per GPU)
  • Total RAM: 23.8 TB
  • Network: 1 Gbit/s Ethernet
  • Central Storage: 122 TB NVMe SSDs
  • Local Scratch Storage: about 70 TB NVMe SSDs

Big Data Analysis Server:

  • Launch: 2023
  • 32 Core EPYC 7513
  • 2 TB RAM
  • 100 GBit/s Infiniband
  • 290 TB net performance optimized NVME SSD RAID

Simulation Cluster 2:

  • Launch: 2021
  • 201 Compute-Nodes
  • 9440 CPU cores 3,0 - 4,1 GHz
  • 42 TB RAM
  • 100 GBit/s Infiniband
  • 170 TB central Storage (120 TB HDD + 50 TB SSD both shared)
  • + 100 TB lokale scratch for temporary data
  • RPeak 912 TFlops

Simulation Cluster 3:

  • Launch: 2016
  • 332 Compute-Nodes
  • 6640 CPU cores  2,2 - 3,4 GHz
  • 42 TB RAM
  • 56 GBit/s Infiniband
  • 180 TB central Storage HDD (138 TB archive + 2*28 TB shared)
  • + 100 TB lokale scratch for temporary data
  • RPeak 235 TFlops (quadrillion [10¹²] floating point operations per second)

Contact

Dr. Harald
Forbert

Laboratory Manager -
Simulation


Room: ZEMOS 0.25
Phone: +49 234 32 - 26751
E-Mail

© ZEMOS
Johannes
Kanacher

Techniker -
IT-System-Administrator, Linux-Server


Room: ZEMOS 01.71
Phone: +49 234 32 - 12169
E-Mail

© Zemos