Kaspar is a GPU computing server based on nVidia Fermi C2070 GPU cards, with a peak floating point performance of 2.6 TFlops.

Specifications

• CPU: dual Xeon X5675, 6 cores/12 threads per processor, 3.06 GHz clock speed, 3.46 GHz Turbo, 6.4GT/s.
• GPU: 4 x NVIDIA Tesla C2070, 448 CUDA cores, 6GB GDDR5 memory, 0.5 TF peak double precision floating point performance.
• Memory: 48 GB (12 x 4GB) 1333 MHz DDR3
• Storage: 2 x 1TB 3Gb/s SATA
• Operating System: CentOS 6 (64 bit)

 

Compilers/Interpreters

• C/C++ : gcc/g++ 4.4, Intel C/C++ XE 12.1 (32 & 64 bit)
• Fortran : gfortran 4.4, Intel Fortran XE 12.1 (includes fortran compiler and Math Kernel Library 10.3 Update 6 - 32 & 64 bit), PGI Fortran compiler v11.7 (64 bit) with CUDA Accelerator for GPU based parallel computing
• Python: v2.6
• Perl: v5.10
• Java: v 1.6 OpenJDK Runtime Environment and 64-Bit Server VM

 

High Performance Computing Tools

• PGI CUDA Fortran v11.7
• Intel Parallel Studio XE 2011, update 2 (Intel C/C++ compiler, Fortran compiler, MKL, libraries & error-checking, security, and profiling tools)
• nVidia CUDA Toolkit 4.1
• pyCuda 2011.2.2

 

Data Analysis/Visualization Packages

• IDL 8.0
  • Setup Instructions: None. IDL is ready to use

  
  

• Mathematica 8.0
  • Setup Instructions: None. Mathematica is ready to use

  
  

• A mailing list, kaspar-users is available for users to discuss software/disk-space requirements or any other relevant topics.
  
  Please contact hpcsupport AT iiap.res.in to log any tickets or requests to install/upgrade software packages.
  

 

Usage Help

• Usage Policies and guidelines
• GPU Computing Tutorials
• GPU Computing Links