Inter-Node | Performance | Network-Based Computing Laboratory

Inter-Node Communication

This page contains sample MPI-level performance numbers for MVAPICH2 using the OSU Benchmarks.

Performance Numbers for MVAPICH2 over OFA-IB-CH3 (05/06/13)
- Intel Sandy Bridge Platform with PCI-Express-Gen3 (Connect IB Dual-FDR)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
- Intel Sandy Bridge Platform with PCI-Express-Gen3 (ConnectX-3 FDR)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks
- Intel Westmere Platform with PCI-Express-Gen3 (ConnectX-2 QDR)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks
- AMD Magny-Cours Platform with PCI-Express-Gen2 (ConnectX-2 QDR)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks
Performance Numbers for MVAPICH2 over uDAPL-CH3 (05/06/13)
- Intel Sandy Bridge Platform with PCI-Express-Gen3 (ConnectX-3 FDR)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks

Performance Numbers for MVAPICH2 over PSM-CH3 (05/06/13)
- Intel Westmere Platform with PCI-Express (InfiniPath-QDR)
  - Two Sided Operations: Latency, Bandwidth and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks

Performance Numbers for MVAPICH2 over OFA-RoCE-CH3 (05/06/13)
- Intel Sandy Bridge Platform with PCI-Express-Gen3 (ConnectX-3 EN)
  - Two Sided Operations: Latency, Bandwidth, and Bidirectional Bandwidth with OSU Benchmarks
  - One Sided Operations: Get, Put and Accumulate Performance with OSU Benchmarks