> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/clara/parabricks/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/clara/parabricks/_mcp/server.

# Frequently Asked Questions

* **What is NVIDIA Parabricks?**

  Parabricks is a software tool for speeding up DNA and RNA analysis.

* **Where does the name come from?**

  We started building small components that could be used to speed up
  computation much like you'd use bricks to build a house.  Those "bricks"
  developed into larger components, which eventually became Parabricks.

* **What hardware and software do I need?**

  For hardware, at least one supported GPU.  How much memory you need depends
  on how many CPU cores you have. Refer to [Installation Requirements](/get-started/installation-requirements) for more
  details.

  For software, Parabricks comes as a Docker container so you'll need the
  NVIDIA Container Toolkit and a Linux OS that supports it.  You'll also need a version of the NVIDIA driver that is compatible with CUDA 12.9 (for example 535, 550, 570, 575, or similar).  Refer to [Installation Requirements](/get-started/installation-requirements) for more details.

  That's all - because Parabricks comes as a Docker container no additional
  software installation is required.

* **How do I ensure that my Parabricks analysis won't terminate if I lose my SSH connection?**

  This is only a problem if you're running on a remote computer - an AWS
  instance, for example.  It's not specific to Parabricks; it can happen with any
  program being run on the remote computer.

  A simple option is to preface your command with [nohup](https://www.computerhope.com/unix/unohup.htm):

  ```sh
  $ nohup docker run ...options... pbrun <Your Command> &
  ```

  This will run your command in the background in a way that ignores all
  "hangup" signals (i.e. loss of SSH connection) and saves all output to a
  file called `nohup.out`.

  Another option is to use a program that supports persistent sessions such as [screen](https://linuxize.com/post/how-to-use-linux-screen/) or [tmux](https://github.com/tmux/tmux/wiki).

* **Parabricks does not run when a system is initialized using Singularity containerization; Singularity does not load all `nvidia` modules.**

  Do this if you see an initialization error:

  ```sh
  $  nvidia-modprobe -u -c=0
  ```

  This is only a concern with versions of Parabricks prior to v4.0.

* **Can I use Parabricks on my video card?**

  It depends on your video card.  Parabricks requires at least 16 GB of GPU memory (38 GB for
  fq2bam, unless you use the `--low-memory` option), and runs on
  'data center' GPUs.  Refer to the [Installation Requirements](/get-started/installation-requirements) for more details.