Cryo-EM GPU nodes

The Center for Research Computing provides ten dedicated GPU nodes on Farnam cluster for Cryo-EM data processing (obtain a Farnam account).


Details on the usage limitation can be found at Farnam website. Expand the ‘PI partitions’ under Private Partitions, then select pi_cryoem or pi_tomography. 

  • For pi_cryoem partition, active Cryo-EM users may apply for access by contacting Shenping Wu or Marc Llaguno. Please provide the user’s netID, PI name and email addresses of the user and the PI. 
  • For pi_tomography partition, active Cryo-EM users may apply for access by contacting Jun Liu. Please provide the user’s netID, PI name and email addresses of the user and the PI. 

List of users that have access to the Cryo-EM GPU nodes (Yale credential required).

These nodes are managed by The Center for Research Computing, please check out the user guides for GPUs and Slurm.  In general, authorized users may use ‘-A cryoem -p pi_cryoem’ or ‘-A tomography -p pi_tomography to specify the resource in your script. 

Due to technical considerations, users may run any type of Cryo-EM data processing, including single particle and tomography, on either partition, regardless of what the name implies. However, each user can only request access to one of them, not both. Please decide wisely. Users may switch from one to another if their needs change. 

These nodes are reserved for jobs that are pertinent to Cryo-EM projects, including image processing and molecular modeling that benefit from running on the GPUs. Please refrain from running jobs that are irrelevant to Cryo-EM. Violators may be suspended. As users of a shared resource, everyone needs to be considerate to others. Here is an incomplete list of good practices:

1. Request only the appropriate amount resource;

2. Release the resource when you no longer need it;

3. Actively check your jobs and avoid letting defunct jobs occupy the resource.

If you suspect there are inappropriate activities on these nodes, you may look up the active user list and send the owner of the job a reminder. You may also report such cases to Shenping Wu or Marc Llaguno, who can kill the jobs.