D-Robotics RDK S100 Robotics Development Kit 80 TOPS

OpenELAB Technology Ltd. Verified independent store

→

The D-Robotics RDK S100 is a high-performance robot development platform built on D-Robotics' proprietary "Dual-Brain" heterogeneous architecture, combining AI perception inference with real-time motion control in a single integrated system. Powered by a Sunrise-class intelligent computing core featuring a BPU Nash neural processing unit delivering 80 TOPS (S100) or 128 TOPS (S100P) of INT8 inference performance, the RDK S100 series represents a significant leap in edge computing for robotics. The "人启双脑" (Human-Inspired Dual-Brain) architecture allocates dedicated processing resources to each domain: the BPU Nash handles AI perception and motion control model inference, while the quad-core Cortex-R7 MCU manages high-frequency motor communication and motion command interpolation — eliminating the bottleneck of running time-critical control loops on a general-purpose CPU. This enables truly integrated perception-and-action pipelines for complex robotic systems such as quadruped robots, dexterous manipulators, and autonomous mobile platforms. With 12GB (S100) or 24GB (S100P) LPDDR5 memory, 64GB eMMC storage, an M.2 NVMe expansion slot, dual Gigabit Ethernet, 4× USB 3.0, a 100-pin camera expansion connector, and comprehensive GPIO/MCU expansion headers, the RDK S100 provides the I/O bandwidth and computing headroom demanded by advanced robotic applications. Backed by 200+ open-source algorithms and D-Robotics' mature development ecosystem, the RDK S100 is ready for the next generation of intelligent robots. Features BPU Nash — 80 TOPS / 128 TOPS (INT8) — D-Robotics' latest-generation Brain Processing Unit delivers industry-leading AI inference throughput for complex perception models, motion planning, and multi-modal fusion at the edge "Dual-Brain" Heterogeneous Architecture — Innovative CPU + BPU + MCU co-processing design where the BPU handles perception inference, the 6× Cortex-A55 CPU manages application logic, and the 4× Cortex-R7 MCU handles motion interpolation and high-frequency communication (up to 1kHz+ servo loops) 6× Arm Cortex-A55 with Safety Features — Includes 1 pair of DCLS lockstep cores and 2 Split-Lock capable cores for functional safety compliance in industrial and automotive robotics applications 4× Arm Cortex-R7 Real-Time MCU — Dedicated real-time processor for motion command interpolation, servo bus communication, and time-critical control loops — offloading the CPU and reducing motion latency to microsecond-level precision Arm Mali-G78AE GPU @ 100 GFLOPS — High-performance GPU for 3D visualization, simultaneous localization and mapping (SLAM rendering), and GPU-accelerated compute tasks 12GB / 24GB LPDDR5, 96-bit Wide Bus — Massive memory bandwidth for running large perception models (Transformer, occupancy networks), multi-sensor fusion, and concurrent AI pipelines (S100: 12GB / S100P: 24GB) 64GB eMMC + M.2 NVMe Expansion — Ample onboard storage for OS and models, plus an M.2 Key M slot (PCIe 3.0 x1) for high-speed SSD expansion to handle large datasets and logging Dual Gigabit Ethernet (RJ45 × 2) — Two independent Gigabit ports for simultaneous sensor network connectivity, ROS 2 communication, and high-throughput data streaming 4 × USB 3.0 + USB 2.0 Type-C — High-bandwidth USB 3.0 ports for LiDAR, depth cameras, and external storage; Type-C port for flashing and serial debug of both main and MCU domains 100-Pin Camera Expansion Connector — High-density camera interface for connecting MIPI CSI camera modules, stereo vision systems, or ToF depth sensors via expansion boards Comprehensive Expansion Headers — 40-pin main GPIO + 16-pin MCU GPIO + 100-pin MCU expansion connector + automatic mode header + JTAG debug, providing maximum flexibility for custom robotics integration M.2 Key E Slot for Wireless Expansion — PCIe 3.0 x1 + UART slot for Wi-Fi and Bluetooth module installation, enabling wireless connectivity tailored to your deployment requirements HDMI 1.4 Output — 2K@60fps — Display interface for HMI visualization, real-time inference preview, and debug monitoring 200+ Open-Source Algorithms — Extensive algorithm library covering object detection, human pose estimation, V-SLAM, occupancy prediction, motion planning, and more — all optimized for the BPU Nash architecture Pre-installed Acrylic Protective Case — Ships with a transparent acrylic enclosure for immediate tabletop development and prototyping Specifications Variant Comparison Parameter RDK S100 RDK S100P NPU (BPU) 1 × BPU Nash, 80 TOPS INT8 1 × BPU Nash, 128 TOPS INT8 RAM 12GB LPDDR5, 96-bit 24GB LPDDR5, 96-bit CPU 6 × Arm Cortex-A55 6 × Arm Cortex-A55 MCU 4 × Arm Cortex-R7 4 × Arm Cortex-R7 GPU Arm Mali-G78AE @ 100 GFLOPS Arm Mali-G78AE @ 100 GFLOPS Storage 64GB eMMC + M.2 Key M 64GB eMMC + M.2 Key M Interfaces Identical Identical Best For Standard robotics, perception Large models, multi-modal AI, complex systems Processor Architecture Parameter Specification AI SoC D-Robotics Sunrise-class Intelligent Computing Core NPU (BPU) 1 × BPU Nash, 80 TOPS (S100) / 128 TOPS (S100P), INT8 quantization CPU 6 × Arm Cortex-A55 (1× DCLS lockstep pair + 2× Split-Lock capable cores) MCU (Real-Time) 4 × Arm Cortex-R7 (motion interpolation, high-frequency communication) GPU 1 × Arm Mali-G78AE, 100 GFLOPS Architecture "Dual-Brain" — BPU (perception) + MCU (motion) heterogeneous co-processing Memory & Storage Parameter Specification RAM 12GB / 24GB LPDDR5, 96-bit bus width Onboard Storage 64GB eMMC Storage Expansion 1 × M.2 Key M slot (PCIe 3.0 x1, NVMe SSD) Connectivity Parameter Specification Ethernet 2 × RJ45, 1000Mbps Gigabit Ethernet Wireless Expansion 1 × M.2 Key E slot (PCIe 3.0 x1 + UART, for Wi-Fi / BT module) USB & Display Parameter Specification USB 3.0 Host 4 × USB 3.0 Type-A USB 2.0 Type-C 1 × Micro USB 2.0 Type-C (flashing & debug, main + MCU domain) HDMI Output 1 × HDMI 1.4, up to 2K @ 60fps Expansion & Debug Parameter Specification 40-Pin Main GPIO Header General-purpose I/O for sensors and peripherals 16-Pin MCU GPIO Header Dedicated MCU I/O for motor/servo interfaces 100-Pin MCU Expansion Connector High-density MCU expansion for custom daughterboards 100-Pin Camera Expansion Camera module interface (MIPI CSI via expansion boards) MCU Port Expansion Header Additional MCU I/O breakout Automatic Mode Header For autonomous mode configuration JTAG Debug Header 1 × JTAG (separate groups for main domain and MCU domain) DIP Pin Function Switch 1 × group, for configurable pin multiplexing RTC Battery Holder 1 × CR-type RTC battery socket Fan Connector 1 × onboard fan power header Power & Physical Parameter Specification Input Voltage 12V ~ 20V DC Rated Voltage 20V Rated Current 10A max Power Connector DC barrel jack (6mm OD × 2.5mm ID) Recommended PSU (Typical) 70W (12V @ 6A) Recommended PSU (Peak) 150W (15V @ 10A) Dimensions (with case) 120 × 121 × 51 mm Weight (with case + packaging) ~126 g Enclosure Pre-installed transparent acrylic protective case RDK S100 vs RDK X5 vs RDK X3 — Series Comparison Parameter RDK X3 RDK X5 RDK S100 RDK S100P Positioning Entry-Level AI Mainstream AI High-End Robotics Flagship Robotics NPU (BPU) 5 TOPS 10 TOPS 80 TOPS 128 TOPS CPU 4× A53 @ 1.5GHz 8× A55 @ 1.8GHz 6× A55 (lockstep) 6× A55 (lockstep) MCU No No 4× Cortex-R7 4× Cortex-R7 GPU No No Mali-G78AE 100GFLOPS Mali-G78AE 100GFLOPS RAM 4GB LPDDR4 4/8GB LPDDR4x 12GB LPDDR5 24GB LPDDR5 Storage TF Card NAND + TF Card 64GB eMMC + M.2 64GB eMMC + M.2 USB 3.0 1 × Type-A 4 × Type-A 4 × Type-A 4 × Type-A Ethernet 1 × GbE 1 × GbE (PoE) 2 × GbE 2 × GbE Power 5V/3A (15W) 5V/5A (25W) 12~20V (70~150W) 12~20V (70~150W) Size 85×56×20mm 85×56×20mm 120×121×51mm 120×121×51mm Reference Documents: RDK S100 Quick Start Guide (Official Wiki) D-Robotics Developer Community — Documentation & SDK Downloads RDK Model Zoo — Pre-trained AI Models (GitHub) RDK S100 Product Page (Waveshare) Application Quadruped & Humanoid Robot Locomotion — The dedicated Cortex-R7 MCU handles high-frequency servo communication and motion interpolation while the BPU runs terrain perception and gait planning models, enabling dynamic locomotion on complex terrain Dexterous Manipulation & Grasping — Run complex manipulation policies (Transformer-based, reinforcement learning) on the 80/128 TOPS BPU with real-time joint control via the MCU domain, achieving microsecond-level motion precision Autonomous Mobile Robot (AMR) Navigation — Dual Gigabit Ethernet enables simultaneous LiDAR and camera connectivity; the BPU accelerates 3D perception, occupancy prediction, and path planning for warehouse and factory logistics Multi-Sensor Fusion & 3D SLAM — Mali-G78AE GPU accelerates map rendering while the BPU processes visual and LiDAR data for real-time SLAM in large-scale environments Edge LLM & Embodied AI — 12GB (S100) or 24GB (S100P) LPDDR5 enables deployment of large language models and vision-language models for natural language robot interaction and task planning Industrial Inspection & Safety-Critical Systems — DCLS lockstep Cortex-A55 cores and Split-Lock mode provide functional safety compliance for industrial robotics and automotive applications Robot Soccer & Competition Robotics — 200+ open-source algorithms include humanoid robot soccer control, enabling rapid development for RoboCup and similar competitions Research & Academic Robotics Labs — Complete development ecosystem with hardware-software co-design, cloud integration, and extensive documentation for advanced robotics research Packing List # Item Details 1 RDK S100 / S100P Development Board Pre-installed with transparent acrylic protective case 2 Power Adapter DC power supply (CN plug, compatible with US plug), 12V~20V output FAQ Q1: What is the difference between RDK S100 and RDK S100P?The two models share identical interfaces, physical dimensions, and architecture. The key differences are: RDK S100 delivers 80 TOPS NPU inference with 12GB LPDDR5 memory, while RDK S100P provides 128 TOPS (60% more AI throughput) with 24GB LPDDR5 (double the memory). Choose the S100 for standard robotics perception and control workloads. Choose the S100P for large-model deployment (LLM, Vision-Language Models), complex multi-modal AI systems, or applications requiring maximum memory bandwidth and inference throughput. Q2: What is the "Dual-Brain" architecture and why does it matter?Unlike conventional SBCs where a single CPU handles everything, the RDK S100's "人启双脑" (Human-Inspired Dual-Brain) architecture assigns dedicated processors to distinct domains: the BPU Nash runs AI perception and motion control models, while the 4× Cortex-R7 MCU handles real-time motion command interpolation and high-frequency servo bus communication. This separation eliminates the latency bottleneck of running time-critical motor control on a general-purpose CPU, achieving microsecond-level motion precision — critical for dynamic legged locomotion and dexterous manipulation. Q3: How does the RDK S100 compare to RDK X5 and RDK X3?The RDK S100 is positioned as the high-end robotics platform, significantly more powerful than the X3 (5 TOPS) and X5 (10 TOPS). Key upgrades include: 80/128 TOPS BPU Nash (vs 5/10 TOPS), dedicated Cortex-R7 MCU for real-time motion control (X3/X5 have none), Mali-G78AE GPU (X3/X5 have none), 12/24GB LPDDR5 (vs 4/8GB LPDDR4x), 64GB eMMC + M.2 NVMe (vs NAND Flash / TF Card), dual Gigabit Ethernet (vs single), and functional safety cores (DCLS lockstep). The S100 targets advanced robotics (quadrupeds, humanoids, manipulators), while X3/X5 are optimized for general AI perception and edge inference. Q4: What operating system and development tools are supported?The RDK S100 runs a Linux-based operating system maintained by D-Robotics. Development is supported through the RDK Studio IDE for application development, NodeHub for one-click algorithm deployment, and RDK Cloud for cloud integration. The platform provides 200+ open-source algorithms covering object detection, V-SLAM, occupancy prediction, motion planning, and more. Official documentation is available at the RDK S100 Quick Start Wiki. Q5: Can I connect Wi-Fi and Bluetooth to the RDK S100?Yes. The board provides an M.2 Key E slot (PCIe 3.0 x1 + UART) designed for installing a Wi-Fi and Bluetooth module. The module is not included by default and must be purchased separately. Compatible M.2 Key E wireless modules (such as Intel AX200/AX210 series) can be installed for Wi-Fi 6 and Bluetooth 5.x connectivity. Q6: What power supply do I need?The RDK S100 accepts 12V ~ 20V DC input via a barrel jack connector (6mm OD × 2.5mm ID). For typical development workloads, a 70W power supply (12V @ 6A) is recommended. For peak loads with multiple cameras, LiDAR, and full AI inference, a 150W power supply (15V @ 10A) is recommended. The included power adapter comes with a CN plug (compatible with US plug); EU/UK users may need a plug adapter or a locally-sourced 12~20V DC power supply. Q7: What cameras and sensors can I connect?The 100-pin camera expansion connector supports MIPI CSI camera modules via expansion boards (sold separately). The 4 × USB 3.0 ports can connect USB cameras, depth sensors (RealSense, Orbbec), and LiDAR units (Livox, RoboSense). The dual Gigabit Ethernet ports support network-attached sensors. The 40-pin GPIO and 16-pin MCU GPIO headers provide I2C, SPI, and UART connections for IMUs, encoders, ToF sensors, and other peripherals. Q8: Does the RDK S100 support ROS 2?Yes. The RDK S100 is designed for robotic applications and supports the D-Robotics TogetheROS.Bot middleware stack, which is compatible with ROS 2. With dual Gigabit Ethernet and extensive expansion headers, it can serve as the central compute node in a ROS 2 robotic system, running perception, planning, and control nodes simultaneously. Q9: What are the DCLS lockstep and Split-Lock CPU features?The RDK S100's 6× Cortex-A55 CPU includes 1 pair of DCLS (Dual-Core Lock-Step) cores that execute the same instructions simultaneously and compare results for error detection — a functional safety mechanism required in ISO 26262 (automotive) and IEC 61508 (industrial) applications. Additionally, 2 cores support Split-Lock mode, allowing them to operate independently (Split) for maximum performance or in lock-step (Lock) for safety-critical redundancy. These features make the RDK S100 suitable for safety-certified robotic deployments. Q10: What is the warranty policy for the RDK S100?Please refer to the manufacturer's warranty policy as specified on the product packaging or official documentation. For warranty claims and technical support, contact the point of purchase or visit the D-Robotics Developer Community. Show more