Compute nodes phase 2

Hardware of Phase II (operating since 2015)

8 login nodes

  • 2x processors „Intel® Xeon® processor E5-2680 v3“
    • Each processor consists of 12 cores
    • Hyperthreading is off
    • There is a AVX2 unit per core (2.1 GHz plus turbo mode)
  • Overall 24 cores per node
  • Processor clock 2,5 GHz – with a turbo up to 3,3 GHz (if you use less cores)
  • Node memory expansion: 128 GByte per node
  • Network connection: 1x FDR-14 InfiniBand and 2x 10GBit-Ethernet
  • Name of the system: hlb0001 … hlb0008
  • Accessible from outside with: lcluster5 … lcluster12

MPI2 – Section (596 nodes)

  • 2 x processors „Intel® Xeon® processor E5-2680 v3“
    • Each processor consists of 12 cores
    • Hyperthreading is off
    • There is a AVX2 unit per core (2.1 GHz plus turbo-mode)
  • Overall 24 cores per node
  • Processor clock is 2,5 GHz – with a turbo up to 3,3 GHz (when using less cores)
  • Node memory expansion: 64 GByte per node
  • Network connection 1x FDR-14 InfiniBand and 1x 1GBit-Ethernet
    • Within the island, all nodes are connected with a blocking factor of 1:1
    • The cross-island blocking factor is approximately 1:8
  • Name of the system: hpb0001 … hpb0596

MEM2 – Section (4 nodes)

  • 4 x processors „Intel® Xeon® processor E7-4890 v2“
    • Each processor consists of 15 cores
    • Hyperthreading is off
    • There is a AVX unit per core
  • Overall 60 cores per node
  • Processor clock is 2,8 GHz – with a turbo up to 3,4 GHz
  • Node memory expansion: 1024 GByte
  • Network connection: 2 x FDR-14 InfiniBand and 1 x 10GBit-Ethernet
  • Name of the system: heb0001 … heb0004

ACC2 (NVD2) – Section (32 nodes)

  • 2 x processors „Intel® Xeon® processor E5-2680 v3“
  • Processors, memory, and network similiar to MPI2 – Section
  • 2x accelerator cards per node
  • 2x nodes with two “NVIDIA® Tesla™ K40m” each
    • each approx. 1.43 TFlop/s computing power (double precision, peak – theoretical)
    • 2880 stream cores
    • 12 GByte main memory
    • 288 GByte/s memory bandwith (peak – theoretical)
    • Name of the system: hab0001 … hab0002
  • 1x node with 2x “NVIDIA® Tesla™ K80 – Dual-GPU”
    • approx. 2,91 TFlop/s computing power (double precision, peak – theoretical)
    • 4992 stream cores
    • 24 GByte main memory (2x12 GByte)
    • 480 GByte/s memory bandwith (peak – theoretical)
    • Name of the system: hab0003
  • The last 29x nodes currently do not have accelerators, therefore the usage is similar to MPI
    • Name of the system: hab0004 … hab0032