Lichtenberg Cluster Darmstadt
Cluster Access
University
Technische Universität Darmstadt
Info
The Lichtenberg Cluster is situated in Darmstadt. It is a tier 2 cluster. Most of the processors are CPU, some accelerators (NVIDIA) are also available.
Note: Phase I of Lichtenberg (Node Type A) is deactivated as of April 2020.
Add. Info
Cluster Introductions:
- Monthly consultation hours for project proposals, in general every first Wednesday a month.
- Monthly introductory courses to the Lichtenberg Cluster, in general every second Tuesday a month.
HPC Support
HKHLR Team Technische Universität Darmstadt
Downloads
Quick Reference Card Lichtenberg Darmstadt
Cluster Access
The cluster is open to researchers from academia and public research facilities in Germany. Access is subject to a scientific project evaluation to the conditions of the steering committee.
Typical Node Parameters
Title
Node type A
Type
Typical
Cores (sockets x cores/socket)
2x8
Memory
32 GB
FLOPS/Core (DP, theor. peak)
20.8 GFLOPS
CPU Type
Intel Xeon E5-2670
MPI Communication (pt2pt)
Intranode
Bandwidth
4.6 GB/s
Latency (64 bytes)
0.73 µs
Memory Bandwidth (triad)
40 GB/s per socket
Accelerators
43 à 2x NVIDIA K20Xm
1.3 TFLOPS, 6GB each
Title
Node type A
Type
Max
Cores (sockets x cores/socket)
8x8
Memory
1 TB
FLOPS/Core (DP, theor. peak)
20.8 GFLOPS
CPU Type
Intel Xeon E7-8837
MPI Communication (pt2pt)
Internode
Bandwidth
4.7 GB/s
Latency (64 bytes)
1.34 µs
Memory Bandwidth (triad)
40 GB/s per socket
Title
Node type B
Type
Typical
Cores (sockets x cores/socket)
2x12
Memory
64 GB
FLOPS/Core (DP, theor. peak)
20.0 GFLOPS
CPU Type
Intel Xeon E5-2680v3
MPI Communication (pt2pt)
Intranode
Bandwidth
4.1 GB/s
Latency (64 bytes)
0.71 µs
Memory Bandwidth (triad)
60 GB/s per socket
Accelerators
2 à 2x NVIDIA K40m, 1 à 2x NVIDIA K80
1.4 TFLOPS, 12 GB each
Title
Node type B
Type
Max
Cores (sockets x cores/socket)
4x15
Memory
1 TB
FLOPS/Core (DP, theor. peak)
20.0 GFLOPS
CPU Type
Intel Xeon E7-4890v2
MPI Communication (pt2pt)
Internode
Bandwidth
6.5 GB/s
Latency (64 bytes)
1.29 µs
Memory Bandwidth (triad)
60 GB/s per socket
Local Tempory Storage
150-300 GB (per node)
Node Allocation
Shared and exclusive
Global Cluster Parameters
Processors (CPU, DP, peak)
771 TFLOPS
Accelerators (GPU, DP, peak)
180 TFLOPS
Computing cores (CPU)
27.928
Permanent Storage
280 TB
Scratch Storage
800 TB
Job Manager
Slurm Workload Manager
Other Job Constraints
runtime: 24h, max. 7d
