Personal tools
You are here: Home → Computing → Hardware → miscellaneous → NVIDIA Tesla GPGPU Computing System
 

NVIDIA Tesla GPGPU Computing System

Hardware Specification

  • node gp01
    • Tesla S2050 GPU Computing System (4 Fermi-based Tesla GPGPUs, 12 GB total memory, ECC)
    • 2 Intel Nehalem quad core CPUs (E5540)
    • 48 GB RAM
  • node gp02
    • Tesla M2090 GPU Computing System (4 Fermi-based Tesla GPGPUs, 22 GB total memory, ECC)
    • 2 Intel Nehalem quad core CPUs (E5540)
    • 48 GB RAM

 

Software Environment

Operating System

  • SLES11 SP2

NVIDIA Driver Version

  • 295.40

CUDA

  • To access the CUDA toolkit, use the command module load cuda which will set the appropriate environment variables for the default CUDA installation.
  • The command module available shows other (non-default) CUDA installations.
  • Currently, the versions 4.0 and 4.1 (default) of the CUDA toolkit are available.

 

Access

The system is reserved for GPGPU development, testing and pilot applications.
The server names are gp01.opt.rzg.mpg.de and gp02.opt.rzg.mpg.de.
Contact: markus.rampp@rzg.mpg.de

 

News

  • Apr 17, 2012: Hardware reconfiguration (gp02 - M2090) and driver updates
  • Jan 30, 2012: CUDA 4.1 and a driver update were installed
  • Apr 29, 2011: CUDA 4.0 was installed

 

Further Information

 

Document Actions