7nm / 12nm / 16nm / 22nm / 28nm
0.2 – 16 TOPS
Ultra-Low Power
Industrial Grade −40°C~+105°C
QADS AI Edge SoC Platform
AQUARIUS EDGE × FUTURE IC DESIGN OFFICE · Scalable · Modular · Power-Efficient · Secure
A scalable, modular AI Edge SoC platform family covering four product tiers — from Nano IoT sensors to Max Edge AI Box — with power envelopes from under 100mW to 5W. Built on 7nm to 28nm process nodes with Cortex-M33 / RISC-V CPU, on-chip NPU, and comprehensive security-by-design architecture.
Product Family & Tier Specifications
| Tier | Target Use Case | CPU | NPU (TOPS) |
Memory | Power | Process Node |
| Nano |
Sensors / IoT |
M0+ | 0.2 – 0.5 | 32 – 64MB | <100 mW | 22 / 28nm |
| Lite |
Smart Camera |
M33 | 0.5 – 1 | 128MB | <500 mW | 22 / 16nm |
| Pro |
IPC / Vision AI |
M33/M55 | 2 – 4 | 256MB | <1.5W | 12 / 16nm |
| Max |
Edge AI Box |
RISC-V / M7 | 8 – 16 | 512MB+ | <5W | 7 / 12nm |
* Process node figures are indicative targets for engineering planning; confirm against final foundry agreements during detailed design.
Top-Level SoC Architecture
CPU TILE
Cortex-M33 / RISC-V
I-Cache 32KB · D-Cache 32KB
FPU · DSP · MPU · NVIC · TrustZone
AI TILE (QADS-NPU)
Scalable 0.2 – 16 TOPS
Tensor Engine · MAC Array
Layer Fusion · Sparse Engine · On-chip SRAM
MEMORY TILE
LPDDR4/4X · DDR4 Controller
SRAM (System) · SRAM (Scratchpad)
Cache Hierarchy (L1/L2) · DMA Controller
VIDEO TILE
ISP Pipeline: HDR·NR·AWB·AE
MIPI CSI-2 RX (4L/4×)
Video Encoder JPEG/H.264/H.265
Frame Buffer Compression
CONNECTIVITY TILE
Wi-Fi 4/5/6 MAC
BLE 5.x · Ethernet MAC (opt.)
USB 2.0 (optional)
SECURITY TILE (ROOT OF TRUST)
Secure Boot ROM · HW Root Key
AES-256/SHA-2 · ECC/RSA Accel.
TRNG · eFuse/OTP · Secure OTA
Q-NOC / AXI FABRIC:
AXI4 Interconnect · QoS & Traffic Shaper · Arbiter/Prioritizer · Bandwidth Manager · Firewall/Isolation
Key Specifications (Summary)
| CPU | ARM Cortex-M33 / RISC-V RV32/64 |
| NPU | 0.2 ~ 16 TOPS (INT8 / INT16 / FP16) |
| Memory | LPDDR4/4X, DDR4, SRAM, Cache L1/L2 |
| Connectivity | Wi-Fi 4/5/6, BLE 5.x, Ethernet (opt.), USB 2.0 (opt.) |
| Video | MIPI CSI-2, ISP, JPEG, H.264/H.265, Compression |
| Security | Secure Boot, AES-256, SHA-2, TRNG, eFuse, OTA |
| Process Node | 22nm / 16nm / 12nm / 7nm (Scalable) |
| Package | BGA / WLCSP / SiP (configurable) |
| Temperature | −40°C ~ +105°C (Industrial Grade) |
POWER ARCHITECTURE
• AON – Always-On Domain
• PD_COMPUTE – Compute Domain
• PD_AI – AI Domain
• PD_MM – Multimedia Domain
• PD_RF – RF Domain
| Deep Sleep | <150 µA (AON only) |
| Wake | <10 mA (Sensor trigger) |
| AI Active | 50–300 mA (NPU running) |
| Full System | 1–1.5W (AI + Wi-Fi TX Peak) |
Typical Data Flow — AI Camera
📷
Image Sensor
(MIPI CSI-2)
→
🔆
ISP
Pipeline
→
🧠
NPU
Inference
→
🎬
Video
Encoder
→
📡
Wi-Fi/BLE
Streaming
→
☁️
Cloud/
Edge Server
SOFTWARE PLATFORM STACK
Application Layer (Customer App / AI App)
AI SDK (QADS AI Runtime) / TFLite / ONNX Runtime / Media Framework
FreeRTOS / Linux / RTOS Kernel
HAL / Driver Layer (BSP) · QADS Silicon Platform
Development & Tools / Engagement
DEVELOPMENT TOOLS
AI Model Compiler & Quantization Toolchain · Debugger/Profiler/Trace
SDK / API / Example Code · Simulation (QEMU)
FPGA Prototyping & Emulation
ENGAGEMENT & DELIVERY MODELS
IP Licensing
License platform IP; customer team takes derivative to tape-out.
Hybrid (IP + Design Service)
QADS provides IP and design services to customize and tape-out the derivative.
Alliance-Built Integration
Integrate partner or licensed IP blocks to accelerate customized derivatives.
APPLICATION EXAMPLES
📷Smart Camera
🏭Industrial IPC
📊Retail Analytics
🌐Edge AI Gateway
⌚Wearable Device
