PC² Benchmark Center

The Benchmarking Center of the PC² offers several services to benchmark high performance (parallel) computer systems. Especially high-performance computing (HPC) systems designed for technical and scientific applications are in the focus of our investigations. Typically, these are cluster systems built out of standard components. Whole systems or components of these systems are evaluated.

• HPC systems / Cluster systems
• High-end Personal Computers and Workstations
• Shared Memory Multiprocessor Systems (SMPs)
• High Speed Interconnects
• Acceleration systems (FPGAs, GPUs, Cell BE, ClearSpeed, ...)

The systems are evaluated on different levels of abstraction:

• Application level: This could be e.g. your application
• System / kernel level: Standard benchmark programs / suites
• Low level; Specialized benchmarks

Have a look at a collection of benchmark results of some test systems.

• Low level benchmarks:
  • Threading Performance
  • Low level InfiniBand performance
  • Low level MyriNet performance
• System level benchmarks:
  • MPI performance of interconnects
  • TCP netperf performance
• Kernel level benchmarks:
  • SparseBench
  • Fast Fourier Transform
  • NAS Parallel Benchmarks
  • High Performance Computing Linpack Benchmark
• Application level benchmarks:
  Will be published soon.

Contact:

Dr. Jens Simon