Skip to content

Frequently Asked Questions

Known Issues

PHENIX/Coot GUI Integration isn't working, is this an issue with my installation?

Answer: If you are using PHENIX 2.0-5936 and coot 1.x, this is not an issue with your setup. For coot, the 0.9.x to 1.x update involved a change from python2 to python3, as did the PHENIX 1.21.x to 2.x update. Both of these changes involved a significant amount of effort from their respective development teams, unfortunately this requires additional effort to restore GUI integration. Using coot to open models, maps, and reflection files from PHENIX still works, as does models from coot with PHENIX.

If you are on a platform where using the older version is an option (which should be all supported operation systems/distributions, aside from macOS 26 Tahoe where coot 1.x is required), you can use a version override to previous versions.

Platforms & Requirements

Will this work on Windows?

Answer: SBGrid is not natively supported on Windows. While some users have success using Windows Subsystem for Linux (WSL), we do not perform internal testing on this platform and cannot guarantee full functionality. More information on supported platforms.

Will this work on other Linux distributions?

Answer: SBGrid generally functions on standard 64-bit Intel Linux distributions. However, our official build and testing environment is limited to CentOS/RHEL-based distributions. View Linux distribution guide.

Hardware and Operating System Requirements

Answer: Requirements are highly dependent on your specific workflow. Linux is the preferred environment for compute-intensive tasks. Apple workstations are supported for general use but are not recommended for CUDA-based CryoEM processing due to hardware limitations. Full hardware specifications.

macOS Support

Issues after upgrading macOS

Answer: Major OS updates can occasionally disrupt the SBGrid environment—most commonly causing the /programs/ directory to disappear or preventing GUI applications from launching. These issues are typically resolved by updating /etc/synthetic.conf or re-installing XQuartz. See more.

Linux Support

Visualization and fonts issues (missing libraries)

Answer: These are usually provided with the graphics card drivers, but in the strange case they are missing check this page.

GPU & Computing

What do I need to run CUDA-accelerated programs?

Answer

  1. NVIDIA GPU Hardware installed in the system.
  2. NVIDIA Drivers: Your Linux workstation must have the appropriate hardware drivers installed.
  3. A version of the title (the executable to run) linked to appropriate CUDA libraries; this includes GPU code (cubin, ptx or both - see below) and CPU (elf) code.
  4. In some cases, the NVIDIA CUDA toolkit is also required.

Do I need to install the NVIDIA CUDA toolkit?

Answer: This depends on the GPU architecture, software title, and version. Here, cubin refers to GPU code for a particular architecture, and ptx refers to intermediate code that can be compiled to appropriate cubin at runtime if necessary. For cases where the version of the title you are using includes cubin matching the GPU architecture you are running on, it is not necessary. If this is not the case (one frequent example being running older titles on newer GPU architectures), the CUDA toolkit is required to generate cubin from ptx for execution. Due to licensing restrictions, SBGrid cannot distribute this toolkit; it must be obtained from a local installation of the NVIDIA CUDA Toolkit. How to configure local CUDA tools for SBGrid.

New Blackwell GPUs support in SBGrid (CUDA>=12.8)

Answer: SBGrid is actively working to support the latest NVIDIA Blackwell GPUs by providing software builds compiled with CUDA >= 12.8 where upstream applications offer native compatibility. Recent additions include packages such as RELION 5.1, AreTomo3, and Xmipp5, all of which include built-in support for Blackwell architectures. With access to Blackwell GPU hardware, SBGrid is now able to test these builds directly and provide official support for compatible applications. In general, SBGrid aims to provide CUDA >= 12.8 versions whenever supported by the original developers; however, we do not extend official support to CUDA >= 12.8 for software that does not yet offer upstream compatibility. As always, user feedback is welcome—please submit a support ticket if you have specific software requests or requirements related to Blackwell GPUs.

Learning & Resources

Where can I find teaching and command line resources?

Answer:

How do I report a bug?

Answer: To ensure a fast resolution, please use the SBGrid web form or email bugs@sbgrid.org. When reporting, please include the output of the sbinfo command from your terminal to help us diagnose the environment. What to include in a bug report.

Still need assistance?

If your question wasn't answered here, please visit our full Troubleshooting Guide for deeper technical diagnostics. For direct assistance, you can always Open a Support Ticket or email our team at bugs@sbgrid.org. To help us help you faster, please include the output of the sbinfo command from your terminal.