AI Gateway: NVIDIA AGX Orin
The Neousys NRU-220S with NVIDIA Jetson AGX Orin provides edge AI capabilities for EQ deployments requiring on-premises intelligence. The fanless industrial system combines gateway functionality with GPU-accelerated processing.
When to Choose AGX Orin
- On-premises AI: GPU-accelerated analytics without cloud dependency
- Multi-sensor gateway: Up to 5 EQ Wave sensors per unit via direct Ethernet connection (6 ports; 1 reserved for facility LAN). Multi-sensor configurations available upon request.
- Air-gapped environments: Defense, critical infrastructure, or facilities with strict network isolation
For deployments without AI requirements, the EQ Gateway provides a simpler, lower-cost solution.
Platform Specifications (Neousys NRU-220S)
| Component | Specification |
|---|---|
| Module | NVIDIA Jetson AGX Orin 64GB |
| AI Performance | 275 TOPS (INT8) |
| GPU | 2048-core NVIDIA Ampere with 64 Tensor Cores |
| CPU | 12-core Arm Cortex-A78AE @ 2.2 GHz |
| Memory | 64GB LPDDR5 (unified CPU/GPU) |
| Storage | NVMe SSD (1-4TB recommended) |
| Ethernet | 2x 2.5GbE (Intel I225) + 4x GbE with PoE+ (802.3at, 25.5W per port, 100W total) |
| Power | 15-60W (configurable power profiles) |
| Operating Temp | -25°C to 50°C |
| Form Factor | Fanless industrial system |
Network Configuration
The NRU-220S has six Ethernet ports with screw-lock connectors:
| Ports | Type | Use |
|---|---|---|
| Port 1 (2.5GbE) | Facility LAN | DHCP or static IP for EQ Sight access |
| Port 2 (2.5GbE) | Available | Second LAN or additional sensor |
| Ports 3-6 (GbE PoE+) | Sensor network | One port per EQ Wave media converter |
For installations with more sensors than available ports, use a managed switch on the sensor subnet.
See LAN Configuration for detailed network setup.
AI Capabilities
When EQ Syntropy is enabled on the AGX Orin, the GPU provides hardware acceleration for physics-informed analytics, event detection, and diagnostic models. AI inference runs locally with no per-query costs, no usage limits, and no external connectivity required. AI capabilities are accessed through EQ Sight.
See EQ Syntropy for details on AI features.
Software Stack
| Layer | Component | Purpose |
|---|---|---|
| OS | JetPack 6.x (Ubuntu-based) | NVIDIA-optimized Linux distribution |
| Runtime | CUDA 12.x, TensorRT | GPU acceleration framework |
| Watch | EQ Watch | Data collection, storage, REST API |
| AI | EQ Syntropy | Physics-informed analytics |
| Interface | EQ Sight | Web-based visualization |
Storage Requirements
CPOW data storage scales with sensor count and retention period:
| Sensors | Daily Storage | 1-Month Retention | 90-Day Retention |
|---|---|---|---|
| 1 | ~50 GB | 1.5 TB | 4.5 TB |
| 3 | ~150 GB | 4.5 TB | 13.5 TB |
| 5 | ~250 GB | 7.5 TB | 22.5 TB |
Storage varies with channel count and signal complexity. The figures above assume a typical 3-phase deployment (~50 GB/day per sensor). Lighter configurations (fewer active channels) use less.
NVMe SSD recommended for primary storage. See Storage Media for capacity planning.
Installation
EQ provides pre-configured AGX Orin systems with all software installed. On-site installation involves:
- Mount the NRU-220S and connect power
- Connect facility LAN to Port 1
- Connect EQ Wave sensors via media converters to the GbE PoE+ ports
- Access EQ Sight via web browser
Contact support@eq.systems for deployment assistance.