Compute Nodes

Nodes Configuration

Salomon is cluster of x86-64 Intel based nodes. The cluster contains two types of compute nodes of the same processor type and memory size. Compute nodes with MIC accelerator contains two Intel Xeon Phi 7120P accelerators.

More about schematic representation of the Salomon cluster compute nodes IB topology.

Compute Nodes Without Accelerator

  • codename "grafton"
  • 576 nodes
  • 13 824 cores in total
  • two Intel Xeon E5-2680v3, 12-core, 2.5 GHz processors per node
  • 128 GB of physical memory per node

cn_m_cell

Compute Nodes With MIC Accelerator

  • codename "perrin"
  • 432 nodes
  • 10 368 cores in total
  • two Intel Xeon E5-2680v3, 12-core, 2.5 GHz processors per node
  • 128 GB of physical memory per node
  • MIC accelerator 2 x Intel Xeon Phi 7120P per node, 61-cores, 16 GB per accelerator

cn_mic

(source Silicon Graphics International Corp.)

cn_mic

Uv 2000

  • codename "UV2000"
  • 1 node
  • 112 cores in total
  • 14 x Intel Xeon E5-4627v2, 8-core, 3.3 GHz processors, in 14 NUMA nodes
  • 3328 GB of physical memory per node
  • 1 x NVIDIA GM200 (GeForce GTX TITAN X), 12 GB RAM

Compute Nodes Summary

Node type Count Memory Cores
Nodes without accelerator 576 128 GB 24 @ 2.5GHz
Nodes with MIC accelerator 432 128 GB, MIC 32GB 24 @ 2.5GHz, MIC 61 @ 1.238 GHz
UV2000 SMP node 1 3328GB 112 @ 3.3GHz

Processor Architecture

Salomon is equipped with Intel Xeon processors Intel Xeon E5-2680v3. Processors support Advanced Vector Extensions 2.0 (AVX2) 256-bit instruction set.

Intel Xeon E5-2680v3 Processor

  • 12-core
  • speed: 2.5 GHz, up to 3.3 GHz using Turbo Boost Technology
  • peak performance: 19.2 GFLOP/s per core
  • caches:
  • IntelĀ® Smart Cache: 30 MB
  • memory bandwidth at the level of the processor: 68 GB/s

MIC Accelerator Intel Xeon Phi 7120P Processor

  • 61-core
  • speed: 1.238 GHz, up to 1.333 GHz using Turbo Boost Technology
  • peak performance: 18.4 GFLOP/s per core
  • caches:
  • L2: 30.5 MB
  • memory bandwidth at the level of the processor: 352 GB/s

Memory Architecture

Memory is equally distributed across all CPUs and cores for optimal performance. Memory is composed of memory modules of the same size and evenly distributed across all memory controllers and memory channels.

Compute Node Without Accelerator

  • 2 sockets
  • Memory Controllers are integrated into processors.
  • 8 DDR4 DIMMs per node
  • 4 DDR4 DIMMs per CPU
  • 1 DDR4 DIMMs per channel
  • Populated memory: 8 x 16 GB DDR4 DIMM >2133 MHz

Compute Node With MIC Accelerator

2 sockets Memory Controllers are integrated into processors.

  • 8 DDR4 DIMMs per node
  • 4 DDR4 DIMMs per CPU
  • 1 DDR4 DIMMs per channel

Populated memory: 8 x 16 GB DDR4 DIMM 2133 MHz MIC Accelerator Intel Xeon Phi 7120P Processor

  • 2 sockets
  • Memory Controllers are are connected via an Interprocessor Network (IPN) ring.
  • 16 GDDR5 DIMMs per node
  • 8 GDDR5 DIMMs per CPU
  • 2 GDDR5 DIMMs per channel