> 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. # Installation Requirements Before you begin running NVIDIA Parabricks, ensure the following hardware and software requirements are met. ## Hardware Requirements The following are hardware requirements for running Parabricks: * Any NVIDIA GPU that supports CUDA architecture 75, 80, 86, 89, 90, 100, 103, 120, or 121 and has at least 16 GB of GPU RAM. NVIDIA Parabricks has been tested on the following NVIDIA GPUs: * T4 * A10, A30, A40, A100, A6000 * L4, L40 * H100, H200 * GH200 * B200, B300 * GB200, GB300, DGX Station with GB300 * RTX PRO 6000 Blackwell Server Edition, RTX PRO 4500 Blackwell * NVIDIA GB10 (DGX Spark) * All tools require at least 16 GB of GPU memory per GPU. The following tools have defaults that require more than 16 GB but have options to lower the device memory usage to fit on devices with 16 GB of GPU memory. * [**giraffe (vg giraffe + GATK)**](/tool-reference/tools/giraffe-vg-giraffe-gatk): Default configuration requires at least 40 GB of GPU memory per GPU. To use a 16 GB GPU, use the `--low-memory` option. More details on tuning the configuration for GPUs with various memory sizes can be found in [giraffe -- Useful Options for Performance](/tool-reference/tools/giraffe-vg-giraffe-gatk). * [**haplotypecaller**](/tool-reference/tools/haplotypecaller) and [**mutectcaller**](/tool-reference/tools/mutectcaller): Default configurations require at least 18 GB of GPU memory per GPU. To use a 16 GB GPU, use the `--htvc-low-memory` and `--mutect-low-memory` options, respectively. * [**minimap2**](/tool-reference/tools/minimap-2): Default configuration requires at least 18 GB of GPU memory per GPU. To use a 16 GB GPU, use the `--low-memory` option. * [**rna\_fq2bam**](/tool-reference/tools/rna-fq-2-bam): Default configuration requires at least 24 GB of GPU memory per GPU. To use a GPU with less device memory, use the `--low-memory` option. * System Requirements: * A 2 GPU system should have at least 100GB CPU RAM and at least 24 CPU threads. * A 4 GPU system should have at least 196GB CPU RAM and at least 32 CPU threads. * A 8 GPU system should have at least 392GB CPU RAM and at least 48 CPU threads. - Running Parabricks on a single GPU is supported but not recommended. If using a single GPU, ensure that your system meets the 2-GPU requirement mentioned above. - Parabricks is supported on time-sliced virtual (vGPU) but not on Multi-Instance (MIG) GPUs. ## Software Requirements The following are software requirements for running Parabricks: * Any NVIDIA driver that is compatible with CUDA 12.9.1 (535, 550, 570, 575, or similar). * Any Linux operating system that supports the [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/index.html) * Docker version 20.10 (or higher). Refer to the following pages for more information on supported driver configurations: * [CUDA Toolkit, Driver and Architecture Matrix](https://docs.nvidia.com/datacenter/tesla/drivers/index.html#cuda-toolkit-driver-and-architecture-matrix) * [CUDA Compatibility](https://docs.nvidia.com/deploy/cuda-compatibility/) Parabricks is available as Docker image. For Singularity users, refer to [Singularity and Docker](https://docs.sylabs.io/guides/2.6/user-guide/singularity_and_docker.html) for more information about importing a Docker image into a Singularity Image, ## Verify Hardware and Software Requirements Ensure you verify the following hardware and software requirements before running Parabricks. ### Checking for Available NVIDIA Hardware and Driver To check your NVIDIA hardware and driver version, use the `nvidia-smi` command: ```sh $ nvidia-smi +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 570.86.15 Driver Version: 570.86.15 CUDA Version: 12.9.1 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA RTX 6000 Ada Gene... Off | 00000000:17:00.0 Off | Off | | 30% 28C P8 10W / 300W | 16MiB / 49140MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 3019 G /usr/lib/xorg/Xorg 56MiB | +-----------------------------------------------------------------------------------------+ ``` This shows the following important information: * The NVIDIA driver version is 570.86.15. * The supported CUDA driver API is 12.9.1. * The GPU has 48 GB of memory. ### Checking for Available CPU RAM and Threads To check how much RAM and CPU threads in your machine, run the following: ```sh # To check available memory $ cat /proc/meminfo | grep MemTotal # To check available number of threads $ cat /proc/cpuinfo | grep processor | wc -l ``` ### Checking for NVIDIA Container Toolkit Installation To make sure you have the NVIDIA Container Toolkit installed, run this command: ```sh $ docker run --rm --gpus all nvidia/cuda:12.9.1-base-ubuntu22.04 nvidia-smi ``` When it finishes downloading the container, it will run the `nvidia-smi` command and show you the same output as above. ### Checking for Python Version To get the version of Python you have, enter the following command: ```sh $ python3 --version ``` Make sure it is at least version 3 (for example, 3.6.9 or 3.7).