Cluster Description

From GPU-Users
Jump to: navigation, search

Currently we are testing 3 base machines each with different graphics card configurations.

SuperServer 7046A-HR+F (Bare bones)
Qty (2) Intel L5630 (2.13GHz/12M/40W)
Qty (18) 4GB 1066 DDR3 ECC Memory

The nodes, GPU(s) and SDK locations are as follows:

You can access them by ssh'ing to alpha.vampire, beta.vampire and gamma.vampire from any primary gateway (any gateway accessible to the outside world). These nodes are located only on the internal network and have access to production GPFS.

We plan to purchase about 40 machines of one of the above configurations. The final cluster will also have high-speed networking between nodes.

The test nodes and cluster are housed and managed by ACCRE - Vanderbilt's Advanced Computing Center for Research and Education and home to our non-GPU based cluster.

Compiling code on the nodes:

  • alpha (cuda):
    • cuda -o myprog myprog.cu
  • beta (opencl):
    • using CUDA: g++ -lOpenCL myprog.cpp -o myprog
    • using ATI-stream: g++ -L/opt/ati-stream-sdk-v2.2-lnx64/lib/x86_64 -lOpenCL myprog.cpp -o myprog -I/opt/ati-stream-sdk-v2.2-lnx64/include
  • gamma (cuda):
    • cuda -o myprog myprog.cu