> This page is for version v4.8.0 (default).
> For other versions, use one of these documentation indexes:
> - v4.8.0 (default): https://docs.nvidia.com/clara/parabricks/v4.8.0/llms.txt
> - v4.7.1: https://docs.nvidia.com/clara/parabricks/v4.7.1/llms.txt

> 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.

# NVIDIA Arm CPU Platforms

Parabricks is distributed as a multi-architecture container and is optimized and
supported on NVIDIA Arm-based platforms — from systems powered by the NVIDIA Grace CPU, to
its successor the NVIDIA Vera CPU, to the NVIDIA GB10 Grace Blackwell Superchip in NVIDIA
DGX Spark. These platforms combine breakthrough CPU and GPU performance with a unified
CPU+GPU memory architecture, enabling faster data access, reduced complexity, and
accelerated genomic analysis workflows.

For the supported GPUs, driver versions, and GPU and host memory requirements on these
platforms, refer to
[Installation Requirements](/get-started/installation-requirements).

## How Parabricks Uses the Arm Architecture

All tools in the Parabricks suite leverage the strengths of NVIDIA Arm CPUs to maximize
performance:

* **Vectorized Instructions**: Parabricks tools use the Arm vector extensions available on
  the host, including Neon and SVE2.
* **NVLink-C2C Interconnect**: The high-bandwidth, memory-coherent NVLink-C2C link minimizes
  latency for GPU-CPU data transfers, optimizing hybrid workflows.
* **High CPU Core Density**: The Arm-based Grace CPU provides up to 72 cores per superchip
  and Vera provides 88 cores, reducing CPU bottlenecks in hybrid workloads.

For more information about the tools and pipelines that run on these platforms, refer to the
[Tool Reference](/tool-reference).

## Platform Notes

### DGX Spark

On DGX Spark, memory is a single shared pool: the CPU and the integrated GPU share it
dynamically, rather than a host pool plus a separate device pool. Options that increase
device memory usage therefore reduce the memory available to the host side of the same run. Refer to
[Tool-Specific Performance Guidelines](/get-started/getting-the-best-performance#tool-specific-performance-guidelines).

## Performance Tuning

To achieve optimal performance for all Parabricks tools on the NVIDIA Grace and Vera CPUs we
refer the users and developers to the
[Grace CPU benchmarking guide](https://nvidia.github.io/grace-cpu-benchmarking-guide/platform/).
This guide illustrates recommendations and best practices directly related to NVIDIA Arm
CPUs and helps you realize the best possible performance for your particular system. For
general CPU tuning, refer to
[CPU Performance](/get-started/getting-the-best-performance#cpu-performance).

## Getting Started

* Parabricks is available as a multi-architecture container, using the Docker command below.
  This command works across systems, from NVIDIA Arm systems to x86 nodes:

  ```bash
  $ docker pull nvcr.io/nvidia/clara/clara-parabricks:4.8.0-1
  ```

  After pulling the image, follow these [Tutorials](/tutorials) for getting sample data, running with FQ2BAM,
  running with HaplotypeCaller, and WGS calling.

* For any questions or support, refer to the
  [NVIDIA Parabricks Community](https://forums.developer.nvidia.com/c/healthcare/parabricks/290).
  Join a vibrant community of researchers and experts to exchange ideas, seek assistance,
  and stay updated on the latest developments in genomic data analysis.

* To learn more about the NVIDIA GH200 Grace Hopper Superchip refer to
  [here](https://www.nvidia.com/en-us/data-center/grace-hopper-superchip/).

* To learn more about the NVIDIA GB200 Grace Blackwell Superchip refer to
  [here](https://www.nvidia.com/en-us/data-center/gb200-nvl72/).

* To learn more about the NVIDIA Vera CPU refer to
  [here](https://www.nvidia.com/en-us/data-center/vera-cpu/).

* To learn more about NVIDIA DGX Spark refer to
  [here](https://www.nvidia.com/en-us/products/workstations/dgx-spark/).