Ambient
Scientific

GPX-10 AI Processor

The most versatile processor for on-device, battery-powered AI differentiation, the way you want it.
GPX10 is the best-in-class processor for always-on embedded AI applications for power constrained edge devices. An integrated SoC with 10 DigAntm AI Cores, ultra-low power ADC, custom SRAM and support for numerous peripherals, GPX10 is completely programmable and the most versatile solution for all your edge AI needs.
Processor

GPX10 Architecture

GPX10 was designed to be a standalone solution for on-device AI applications. Its architecture was purpose-built with the anatomy of neural networks in mind while minimizing power consumption and maximizing programmability and flexibility.
Ambient Scientific GPX10 Block Diagram
Highlights

GPX10 Highlights and Features

GPX10 Product Brief
Ultra-Low Power for Always-On AI Applications
Always-On detection, <80μW power usage.Years of Always-On AI on one coin-cell
Runs Always-On AI applications for multiple years on a single coin-cell battery
Charger-less AI applications possible by harvesting energy from various natural sources such as kinetic energy and RF energy
Highly Integrated
Multi-channel ADC
Sensor fusion – Connect up to 10 analog and digital sensors simultaneously
ARM M4F for executing non-AI workload
Sophisticated clocking structure to reduce SoC Power Consumption
Versatile & Fexible
Software-defined Programmable AI cores
Support for multiple operand resolutions for both weights and inputs that can be modified dynamically to optimize performance and power consumption
Each core can be enabled, disabled and halted dynamically
DigAntm AI engine can change behavior between being a master or slave. This capability enables highly optimized, high performance and ultra-low- power AI applications

Download GPX10 Product Brief

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Processor Comparison

GPX10 vs. Competitors

Sensor Fusion AI
Features
Ambient GPX10
ASIC Chip
MCU with built-in NPU
Peak perf (GOPs)
512
0.24
100
Active Power
1
(uW)
40-120
200
10,000
Efficiency
3
(TOPS/Watt)
7.3
1.2
<1
Vision/Imaging AI
Features
Ambient GPX10
ASIC Chip
Typical GPU
Peak perf (GOPs)
512
0.24
500
Active Power
2
(uW)
20-50
300-2,000
6,000
Efficiency
(TOPS/Watt)
7.3
1-3
<1
Footnotes:
1 - For Sensor Fusion AI Applications
2 - For Vision and Imaging AI Applications
3 - Measured in Silicon
Platform

Develop and Scale
Powerful AI Applications

Complete development platform with Chip,
DVK and software toolchain
Chip
GPX10 AI SoC with 10 DigAntm AI Cores, Arm M4F co-processor, built-in ultra-low power ADC, sensor DMA and more.
Read More
Development Kit
Ambient's DVK provides a complete development platform for prototyping and debugging embedded AI applications with support for several peripheral interfaces.
Read More
SDK
Ambient's SDK lets you build your own AI applications using popular frameworks such as Tensorflow and Keras. Our software toolchain includes a compiler and training toolkit so you can focus on owning your AI differentiation.
Read More
Ambient Scientific GPX10 Ambisense Development Board
Development Board

The AmbiSense Prototyping Platform

Own your embedded AI applications
On-board Sensors
Analog and digital
microphone, accelerometer,
gas sensor and many more
Integrated Camera
Digital Image sensor with a
full camera system
Other Peripherals
Bluetooth module for communication,
serial flash for storage and support
for several interfaces
Flexible Design
JTAG port, USB port for terminal connection using UART, etc. for easy debugging
USB Port
USB Port that can be used to power up the board. It can also be used to print on serial terminal via UART Bridge. Overall, can be used for the following functions:
Key Notes
Power
Display
Debug
GPX10 Chip
GPX10 AI SoC with our proprietary AI core that consumes ultra-low power and can enable several always-on AI applications for embedded and battery-powered devices such as:
Key Notes
Wake Work
Sensor Fusion
Facial Recognition , etc.
Bluetooth Communication Module
Microchip Bluetooth Module for communicating data and alerts triggered by AI inference or user programmed events. Our board is designed to be flexible with different communication modules such as:
Key Notes
BLE
Wi-Fi
Zigbee, etc.
Header Pins
Several header pins for connecting network of I2C sensors, external flash devices and power sources and includes:
Key Notes
3 Pin connector for I2C sensors
3 Pin Connector for external flash and SPI1 Master of GPX10
3 Pin Connector for external analog sensors
2 Pin Connector for external coin cell battery
LEDs and Buzzer
Multiple LEDs and Programmable Buzzer for generating alerts upon AI inferences or as per user specifications. These can be utilized in multiple ways such as:
Key Notes
Combination of LED flashing sequences
LED colors
Buzzer frequencies and patterns
HM01B0 Camera Connector
Connector for Ultra-low power Image sensor that can be used for processing visual input to enable various applications such as
Key Notes
Facial Recognition
Driver Monitoring
Number Plate Recognition, etc.
On-Board Sensors
Several I2C devices for sensor fusion and other applications such as fall detection, tilt angle etc. These include
Key Notes
3-axis Accelerometer
3-axis Gyroscope
Magnetometer
Analog Mux
External IC that can used for connecting different analog sensors via MUX logic from GPX10. Commonly used sensors include
Key Notes
Temperature Sensor
Pressure Sensor
Optic Sensor
Gas Sensor, etc.
Microphones
Digital and Analog Microphones for processing audio input to enable user flexibility and tradeoff between power consumption and signal-to-noise ratio. These include:
Key Notes
Knowles Digital I2S Mic
Knowles Analog Mic
SDK

Develop and own your embedded AI models using our robust software toolchain

Request SDK
Develop for Edge/µEdge
Algorithms and libraries suitable for edge devices to run computer vision, sensor fusion and voice/speech related applications with a smaller resource footprint which includes network optimization and pruning to fine tune AI performance.
Optimal Hardware Utilization
Ambient's AI stack gives developers the freedom to select AI cores dynamically for their computation without compromising performance.
Secure your data
Allows users to ensure all AI workloads transact locally without relying on cloud data transfers for security and privacy. Additionally, our encryption and decryption algorithms keep data secure even as it leaves the edge device.

Request SDK

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
AI Stack

Ambient AI Stack

Ambient Scientific AI Software Stack
AI/ML Frameworks
Keras
Tenserflow
ONNX
Compiler Toolchain
ONNX Runtime
MX8 Runtime
ONNX-MLIR
Ambient Libraries
MX8 Math Libraries (Arithmetic Logic, Compare, etc.)
MX8 AI Libraries (Convolution, Recurrent, etc.)
Profilers
Debuggers
Compilers
Emulation
System Software
Peripheral Drivers
System Interface
Emulation Support
Operating System
Windows
RTOS
AI Stack

Ambient AI Stack

AI/ML Frameworks
Keras
Tenserflow
ONNX
Compiler Toolchain
ONNX Runtime
MX8 Runtime
ONNX-MLIR
Ambient Libraries
MX8 Math Libraries (Arithmetic Logic, Compare, etc.)
MX8 AI Libraries (Convolution, Recurrent, etc.)
Profilers
Debuggers
Compilers
Emulation
System Software
Peripheral Drivers
System Interface
Emulation Support
Operating System
Windows
RTOS
Know More

Explore Our Applications And Technology

We're innovating remarkable molecular electronics to digitize biology, revolutionizing scientific research and healthcare.
Get In Touch
Headquarter ( Silicon Valley )
Ambient Scientific Inc.4633 Old Ironsides Drive Santa Clara California 95110. USA
Headquarter ( India )
Ramky House, 1st Cross, Raghavendra Nagar, Kalyan Nagar, Bengaluru Karnataka, 560043, India
Newsletter

Exploring the forefront of cutting-edge chip processor technology?

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
By providing your email, you consent to receive promotional emails from Ambient, and acknowledge our terms & conditions along with our privacy policy.