AE-NJ60BT vs AE-3588BT: NVIDIA Jetson Orin NX vs. Rockchip RK3588J Edge AI Comparison
An engineering performance analysis comparing BITECH’s AE-NJ60BT (157 TOPS Jetson Orin) with the AE-3588BT (58 TOPS Scalable RK3588J) on AI compute architecture, GMSL2/MIPI vision ingress, and industrial toolchains.
Core Engineering Specs: Jetson Compute Architecture vs. Native ARM RISC SoC
AE-NJ60BT
High-performance Jetson Orin platform for multi-camera vision AI, AMR/AGV perception, and workloads that require the NVIDIA CUDA / TensorRT ecosystem.
AE-3588BT
Cost-optimized ARM Edge AI node for field control + lightweight inference, with scalable NPU expansion and a long industrial lifecycle on the RK3588J SOM.
Core Engineering Comparison
AE-NJ60BT Wins on: Pure AI inference power (157 TOPS), GMSL2 long-distance vision, and NVIDIA’s mature CUDA/TensorRT ecosystem.
AE-3588BT Wins on: Cost-per-node, long-term 5-7 year ARM lifecycle stability, and versatile field-bus control via RK3588J native I/O.
| Engineering Factor | AE-NJ60BT | AE-3588BT | Engineer Note |
|---|---|---|---|
SOM / Processor | NVIDIA Jetson Orin Nano / NX / NX Super (8GB / 16GB LPDDR5) | Rockchip RK3588J octa-core (4× A76 + 4× A55), up to 8GB LPDDR4 | Defines AI ecosystem (CUDA vs ARM NN/RKNN). |
AI Compute | Up to 157 TOPS (Orin NX Super 16GB) | 6 TOPS built-in NPU, scalable to 58 TOPS via 2× M.2 NPU | NJ60BT for heavy multi-model vision; 3588BT for control + light AI. |
Camera Ingress | 4-lane MIPI CSI or 4× GMSL2 (via adapter board) | MIPI CSI + USB cameras (no native GMSL2) | GMSL2 enables long-distance, EMI-resistant camera links. |
Ethernet | 4× GbE (1× SOM + 3× Intel i210-AT) | 3× GbE | More LANs allow stricter subnet isolation. |
| CAN FD | 2× CAN FD (isolated) | 2× CAN (isolated) | CAN FD doubles payload and bitrate for modern vehicle/AMR buses. |
| Digital I/O | 8 DI / 8 DO opto-isolated, selectable 3.3V/5V/12V | 4 DI / 4 DO opto-isolated | Higher I/O count reduces external PLC dependency. |
| Video Decode | 1× 8K30 / multi-stream 4K | 8K60 H.265 (RK3588J VPU) | Both handle 4-camera 1080p inference comfortably. |
Industrial Isolation | CAN / DIO / Serial isolated | CAN / DIO / Serial isolated (4000V surge) | Both are field-grade. |
Operating Temp | -40°C to +70°C, fanless | -40°C to +80°C, fanless | Both validated for outdoor / vehicle deployment. |
| Power Input | DC 9–36V wide range | DC 9–36V wide range | Identical — fits AGV/AMR/vehicle batteries directly. |
| Software Stack | NVIDIA JetPack, CUDA, TensorRT, DeepStream | Linux / Android, RKNN Toolkit, ROS2 | Choose based on existing model toolchain. |
| Typical Deployment | Multi-camera AOI, AMR perception, autonomous driving R&D | Edge gateway, lightweight vision, robotics control node | Compute class drives the choice. |
Edge AI Deployment Guidelines: Selecting the Right Node Framework
Choose AE-NJ60BT if:
- •You need ≥ 60 TOPS for multi-model concurrent inference
- •4× synchronized cameras (GMSL2 long-distance) are required
- •The team uses CUDA / TensorRT / DeepStream
- •Application is AMR perception, AOI, or autonomous driving
- •4 isolated GbE subnets are required
Choose AE-3588BT if:
- •AI workload is ≤ 32 TOPS (single model or light multi-model)
- •Cost and power budget are primary constraints
- •The team uses ARM Linux / Android / RKNN
- •Application is field gateway, control node, or HMI + light vision
- •Long lifecycle (5–7 years) on a fully-controlled BOM is critical
Deep Dive Product Data & Hardware Manuals
AE-NJ60BT Hardware Data: 157 TOPS NVIDIA Orin NX Computing Node
Comprehensive pinouts, JetPack installation logs, and GMSL2 carrier board topology.
AE-3588BT Hardware Data: Rugged RK3588J 32 TOPS Energy Gateway Node
Dual isolated CAN Bus telemetry records and 4000V surge validation lab test configurations.