Skip to main content

How is Negishi different than other Community Clusters?

Negishi differs from the previous Community Clusters in several significant aspects:

  • Host naming convention in the Negishi cluster is different from earlier Community Clusters. Everything Negishi-related is contained within a negishi.rcac.purdue.edu subdomain. Front-end login nodes are now named loginNN (as opposed to earlier <cluster>-feNN), and compute nodes of each type X are named xNNN (as opposed to <cluster>-xNNN).
  • Negishi OnDemand Gateway is at the gateway.negishi.rcac.purdue.edu (as opposed to earlier gateway.negishi.rcac.purdue.edu convention).
  • Negishi home directories are entirely separate from other Community Clusters home directories. There is no automatic copying or synchronization between the two. At their discretion, users can copy parts or all of the Community Clusters home directory into Negishi - instructions are provided.
  • Negishi contains the 3rd generation of AMD EPYC processors, codenamed "Milan". These CPUs support AVX2 vector instructions set. When compiling your code, use of -march=znver3 flag (for latest GCC, Clang and AOCC compilers) or -march=core-avx2 (for Intel compilers and GCC prior to 11.0) is recommended.
  • GCC compiler with OpenMPI or MVAPICH2 MPI libraries are recommended for software development on Negishi. You can enable this software with module load gcc openmpi (default) or module load gcc mvapich2.
  • If you use Jupyter notebooks, JupyterHub on Negishi will be available only via the OnDemand Gateway rather than the freestanding version as on some previous systems. Other RCAC systems will transition to OnDemand as well, following Negishi.

Link to section 'Upcoming 2023' of 'How is Negishi different than other Community Clusters?' Upcoming 2023

  • A subset of Negishi compute nodes contain AMD Radeon Instinct MI210 accelerator cards which can significantly improve performance of compute-intensive workloads. These can be utilized by submitting jobs to the gpu queue (add -A gpu to your job submission command).
  • A selection of GPU-enabled ROCm application containers from the AMD InfinityHub collection is installed.
Helpful?

Thanks for letting us know.

Please don't include any personal information in your comment. Maximum character limit is 250.
Characters left: 250
Thanks for your feedback.