Bender – GPU Cluster

The small GPU cluster Bender was purchased and installed by the HRZ in 2022. It is mainly intended to be used for teaching contexts but suitable applications for science projects will also be given access provided that enough resources are available.

server2
© Volker Lannert / Uni Bonn

Technical Details

4 A40 GPU Compute Nodes

  • CPU: 2 × Intel Xeon Silver 4309Y 8-core/16-thread 2.80GHz
  • RAM: 404GB
  • GPU: 4 × Nvidia GA102GL [A40] 48GB
  • Local Storage: 2 × 480GB SSD Raid1
  • Network: 2 × 1G Ethernet Uplink, 1 × 200Gbit Infiniband
  • Total Amount of Cores: 64
  • Total RAM: 1.6TB

2 A100 GPU Compute Nodes

  • CPU: 2 × AMD EPYC 7713 64-core/128-thread 2.00GHz
  • RAM: 2151GB
  • GPU: 8 × Nvidia GA100 [A100] 80GB
  • Total Amount of Cores: 256
  • Total RAM: 4.3TB

1 Login Node

  • CPU: 2 × Intel Xeon Silver 4314 16-core/32-thread 2.40GHz
  • RAM: 810GB
  • Local Storage: 2 × 480GB SSD Raid1
  • Network: 2 × 10G SFTP+ Uplink, 1 × 200Gbit Infiniband

1 Storage Node

  • Scratch Storage: 10 × 4TB HDD Raid6
  • CPU: 2 × Intel Xeon Silver 4309Y 8-core/16-thread 2.80GHz
  • RAM: 128GB
  • Local Storage: 2 × 480GB SSD Raid1
  • Network: 2 × 1G Ethernet Uplink, 1 × 200Gbit Infiniband

Operating System & Software

  • OS: Rocky Linux
  • Batch System: Slurm
  • GPU Parallel Computing: CUDA

Access

Fill out the form below to register for a Bender account.

Contact

For all technical questions / support requests, please contact:
bender-support@hpc.uni-bonn.de

Personal Contact

Avatar Barbi

Dirk Barbi

Avatar Kuckertz

Michael Kuckertz

-

Wird geladen