January 2025 CCR Updates

Release of ccrsoft/2024.04 

This new software release contains updated compilers, toolchains, and many of the most popular software applications used on CCR's systems.   

  • Users may test this out using the command: module load ccrsoft/2024.04 
  • You can use the vortex-future login node where this environment is now the default: vortex-future.ccr.buffalo.edu   
  • The default software environment on the login (vortex.ccr.buffalo.edu), compile, and compute nodes, as well as in OnDemand, remains ccrsoft/2023.01 at this time.  
  • You can pin your account to a specific release following the instructions in the green box on this page: https://docs.ccr.buffalo.edu/en/latest/software/releases/ 

  

NOTE: We will no longer be including machine learning type codes such as Pytorch, Transformers, etc in the CCR software environments.  They change so quickly and are difficult to install with Easybuild, making keeping them up-to-date nearly impossible.  Instead, we recommend using NVIDIA containers for these software packages.  They are portable, up-to-date, and available in many configurations.  You can access the NVIDIA container library here and CCR’s container documentation here.  We've provided an example of a workflow in which you might pull an NVIDIA container and use a virtual environment to install additional packages.  In the new Python course (see below) additional guidance will be provided.

  

  

New compute nodes available in the UB-HPC cluster 
There are new compute nodes in the industry compute pool that are also available in the scavenger partition for academic users.  They include:

  • 2 NVIDIA DGX nodes with 8 H100 GPUs and 2TB of RAM
  • 8 high-bandwidth memory nodes
  • 10 nodes with two NVIDIA L40S GPUs and 1TB of RAM
  • 8 nodes with 2TB of RAM
  • 62 standard compute nodes with 512GB of RAM

You can see the node hardware specs with the following commands: 

snodes all ub-hpc/scavenger |grep CPU-Platinum-8562Y


For instructions on how to use the scavenger partitions, please refer to the documentation here. 

  

Intro to CCR Course 

A new course was published in UB Learns for the Fall 2024 semester to provide new users with background information on how to effectively and efficiently use CCR's HPC resources.  The course was updated based on student feedback and republished for the Spring 2025 semester.  All students taking a course that uses CCR are required to complete the Intro to CCR course first and provide their instructor with a certificate of completion before getting access to their course's allocations.  Anyone using CCR for research purposes is highly encouraged to check out the course. You can self-enroll here. https://ublearns.buffalo.edu/d2l/le/discovery/view/course/285151 

  

  

Using Python at CCR Course 
A new course will soon be available in UB Learns that guides CCR users through the various methods for properly using Python in CCR's HPC environment.  We'll make the link available here when the course is published.  Check back here for updates! 



CCR Migrating to TeamDynamix

You may be aware that UBIT has migrated to the TeamDynamix help desk platform, along with other IT units at UB.  This spring CCR will be joining them!  We'll make sure all CCR users are aware of any impact this will have on them well in advance of the cutover.