Intel Xeon Phi coprocessor

Intel Xeon Phi inside

The Intel® Xeon Phi™ coprocessor optimizes highly parallel computing with its many-integrated-core (MIC) architecture, featuring high memory bandwidth and floating point performance. Intel has dubbed the Intel® Xeon Phi™ coprocessor a “supercomputer in a chip” because it offers one teraflop of performance, the same as the first Intel-based supercomputer built in 1997.

This new coprocessor, which plugs into a PCIe slot and features over 50 processing cores, is Intel’s first MIC product, designed to work in tandem with traditional processors. Since this chip was built on Intel processor architecture, it can be programmed just like a typical multiprocessor system with Intel’s standard development tools, including Cluster Studio XE as well as other cluster tools such as GridEngine, Moab and OpenMP. The Xeon® Phis™ were designed based on Intel’s existing ecosystem of programming languages and tools with the idea that what runs on one processor family will run on the other. Of course, recompiling will need to be done, but the time software developers have to devote to this is greatly reduced as they won’t need to learn new tools or programming methods.

Advanced Clustering’s ACT Utils cluster management suite eases integration of Xeon® Phi™

Since the Xeon Phi™ acts much like a “computer in a computer,” managing and administrating a cluster of them can be quite complex. We’ve extended and enhanced our ACT Utils cluster management software suite to better handle this new coprocessor paradigm.

The newest versions of ACT Utils facilitate easy setup, administration and management of Advanced Clustering systems with Xeon® Phi™ cards. The software supports the latest Intel MPSS software stack, multiple coprocessors per machine and all of the programming models (offload and native execution). The tools facilitate management of the Phi’s micro-OS and negate the need to separately manage users, SSH keys, IP assignment and static bridge creation.

Intel Xeon Phi Coprocessor options

Xeon Phi 3120

  • passive or active cooling
  • 57x 1.1GHz cores
  • 6GB of memory
  • 240 GB/s memory bandwidth
  • 1,003 gigaflop double-precision
  • 300W TDP

Xeon Phi 5110

  • passive cooling
  • 60x 1.053GHz cores
  • 8GB of memory
  • 320 GB/s memory bandwidth
  • 1,011 gigaflop double-precision
  • 225W TDP

Xeon Phi 7120

  • passive cooling
  • 61x 1.238GHz cores
  • 16GB of memory
  • 352 GB/s memory bandwidth
  • 1,220 gigaflop double-precision
  • 300W TDP

Systems that support accelerators (including Xeon Phi)

Menu

Advanced Clustering Technologies