Hardware requirements¶
Check this page before you install. RD Analytics needs a GPU-capable PC to run video analysis.
Linux (Ubuntu) is the recommended platform for the server. Windows 11 with WSL2 is a supported alternative when a Linux host is not available.
Who needs what¶
| Role of this PC | What you need |
|---|---|
| Server (runs RD Analytics) | NVIDIA GPU + driver, enough RAM/disk — see below |
| Client (browser only) | Any modern PC; open http://<server>:9006 |
Server requirements (baseline)¶
| Requirement | Notes |
|---|---|
| Architecture | x86_64 (64-bit) |
| OS | Ubuntu 24.04 (recommended), or Windows 11 with WSL2 Ubuntu (install guide) |
| NVIDIA GPU | Required for the analysis engine |
| NVIDIA driver | 590+ on the host (Windows: install the Windows driver; Linux: install on the host) |
| GPU runtime | Set up automatically by install.sh on Ubuntu when missing |
| RAM | 32 GB minimum practical; 64 GB for Pro / multi-scan hosts |
| Disk | 50 GB+ free at install; plan hundreds of GB–TB for real media archives |
Note
The installer does not install the NVIDIA driver. Install that yourself first, then run the RD Analytics installer.
Sample PC configurations¶
These are example builds to help you choose hardware. Ask your vendor if you need a quote for a specific site.
Minimum / 50fps license¶
Suitable for first installs, demos, and 50fps licenses with mostly one scan at a time.
| Component | Sample configuration |
|---|---|
| OS | Ubuntu 24.04 LTS (recommended), or Windows 11 + WSL2 Ubuntu |
| CPU | Intel Core i5 / AMD Ryzen 5 (6+ cores), recent generation |
| RAM | 32 GB |
| GPU | NVIDIA RTX 3050 |
| Driver | NVIDIA 590+ |
| Storage | 512 GB NVMe SSD (system + /opt/rd_analytics data) |
| Network | Gigabit Ethernet if other PCs open the UI on the LAN |
Expect comfortable single-pipeline file processing within a 50 FPS budget. Heavy models or several parallel scans may feel slow on this class of GPU.
Recommended — Pro license¶
Aimed at Pro (uncapped) licenses where the product does not throttle file processing by FPS budget, and you may run stronger models or more parallel work.
| Component | Sample configuration |
|---|---|
| OS | Ubuntu 24.04 LTS (recommended), or Windows 11 + WSL2 Ubuntu |
| CPU | Intel Core i9+ or AMD Ryzen 9+ (12+ cores) |
| RAM | 64 GB |
| GPU | NVIDIA RTX 5080 / 5090 class, or workstation equivalent |
| Driver | NVIDIA 590+ |
| Storage | 2 TB NVMe SSD (OS + media + compiled engines); add more if you keep large video archives |
| Network | Gigabit (or faster) Ethernet for multi-user / LAN access |
Use this class of machine when you want headroom for Pro’s uncapped budget, larger AI models, and multiple concurrent pipelines. For continuous high-volume sites, size disk for your media retention and confirm GPU choice with your vendor.
Tip
Pro does not replace a fast GPU. An uncapped license removes the software FPS cap; throughput is still limited by GPU, CPU, disk, and how many scans you start together.
Quick checks¶
Linux
nvidia-smi
uname -m # should print x86_64
df -h # confirm free space
Windows 11
- Confirm the GPU appears in Task Manager → Performance (or NVIDIA Control Panel).
- After WSL prep, GPU access is used from Ubuntu inside WSL; keep the Windows NVIDIA driver up to date.
Compiled AI engines¶
The first time you enable an AI model, the engine may compile a GPU engine for your hardware (status COMPILING). That can take several minutes and uses disk under the system directory.