> For the complete documentation index, see [llms.txt](https://nx.docs.scailable.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-manager-plugin/quick-start/system-requirements.md).

# System requirements

**Minimum Hardware Requirements**

* **x86\_64 Systems**: Intel i5 (10th Gen quad-core equivalent) or higher, with at least 4GB of RAM\
  The processor must support AVX2 instructions; check the [list of AVX2-compatible processors](https://avx2checker.com/processors) to confirm.
* **ARM Systems**: Raspberry Pi 4B equivalent or higher (AARCH64), with at least 4GB of RAM
* **NVIDIA Jetson**: An NVIDIA Jetson device with [Jetpack 5, 6 or 7](/ai-manager-v6.1.4/ai-accelerators-support/nvidia-support.md) installed

**OS Requirements**

The AI Manager is supported on the Mediaserver on Ubuntu Linux LTS version 20.04 or higher, Windows 10 and Windows 11.

The AI Manager is not supported on systems running Windows Server 2016 and older versions.

## Performance Recommendations

{% hint style="info" %}
**NOTE**: These recommendations assume the computer runs only the Nx Meta Mediaserver application with default settings and one AI plugin. Running additional applications concurrently may impact performance and stability.
{% endhint %}

For example, a system with one camera, no saved layouts, few rules, and a single user on the minimum required hardware performs adequately. A system with multiple models, many cameras per server, hundreds of layouts, numerous rules, and many concurrent users requires substantially more hardware.

## AI Input Resolution and Secondary Stream Recommendation

Enable the plugin on a camera secondary stream with a resolution of 512x512 pixels or slightly higher (e.g., 720p) and a minimum frame rate of 6 FPS. Refer to the camera's documentation to configure the secondary stream accordingly.

Enabling the AI Manager plugin on a camera primary stream is possible but not recommended. Running the plugin on a primary stream requires the Mediaserver to decode the high-resolution stream and scale it down, which imposes a high CPU load and significantly increases memory consumption. Running AI plugins on the primary stream does not improve detection accuracy and reduces the number of AI detections per second on the Mediaserver.

## AI Detections per Second

For accurate AI detection in security or surveillance use cases, a minimum of 6 detections per second is typically sufficient. By default, the AI Manager plugin performs as many detections per second as possible. The number of detections per second is limited by the number of frames per second in the video stream and the AI processing power of the device it runs on.

On an Intel i5 12th gen device with 8GB of memory, using 720p resolution secondary camera streams, the following performance can be expected:

Single 720p RTSP H264 secondary stream at 24 fps stream performance with Detection plugin:

* Up to 23.2 FPS for the large people detection model.

Four 720p RTSP H264 secondary streams at 24 fps stream performance with Detection plugin:

* Up to 9.0 FPS per stream for the large people detection model

These numbers serve as a reference point for estimating system performance; actual performance may vary based on specific configurations, hardware variations, and other factors.

## Hardware Recommendations

When the AI Manager is enabled on one or multiple cameras of a Mediaserver, that system's CPU/GPU load and memory consumption increase. This is because the Mediaserver needs to decode the camera streams for which AI Manager is enabled, and the AI Manager plugin requires resources to perform the AI inference.

The following recommendations assume that the device is dedicated to Nx Meta with the AI Manager Plugin with default settings and that the AI Manager Plugin is enabled on 720p secondary camera streams.

The requirements are in addition to the basic hardware requirements of Nx Meta, so especially RAM and Storage requirements must be added to those.

## Memory Requirement

Add 512 MB of RAM capacity per stream for which the AI plugin is enabled.

## CPU/GPU Requirement

Up to 6 cameras with AI detection: Intel i5 10th gen or AMD Ryzen 5 3000 Quad-Core.

Up to 12 cameras with AI detection: Intel i7 12th gen or AMD Ryzen 7 3000 Quad-Core.

## Disk Space Requirements

There are no fixed disk space requirements for the AI Manager.

The hardware specific runtimes each have a different size ranging from a few MB to 5GB.

* Nvidia Runtime uses approximately 500MB
* Intel OpenVino Runtime uses approximately 4GB

Additional disk space is needed to store the metadata that is generated from running AI detection. This is disk space that is part of the Mediaserver data storage space.

## Other Hardware

For other hardware architectures such as ARM, NVIDIA Jetson Orin, or Hailo, contact your reseller.