Feature: Edge AI
This implementation of the DigAn Figure 5: Each MX8 core is a self-contained neural network processing unit
power. We are seeing over 100 times better performance than a typical MCU with the same power consumption, or the same performance as a typical GPU but using less than 1% of the energy.
New silicon technology realised in GPX family chips This breakthrough in power and performance is the result of Ambient Scientific’s decision to embrace a new
compute architecture and to create new silicon technology to implement it. For the embedded system developer
who wants to enable high-speed, low- power AI, however, this fundamental technology needs to be practically accessible; design engineering teams need the technology in chip form, supported by a development environment for porting trained AI models to the device.
technology is available in the form of the GPX family of chips, of which the first is the GPX10, succeeded by the GPX10 Pro later in 2025. In the GPX products, DigAn blocks are assembled into AI processor cores, named MX8, which can be scaled up to meet different application needs; see Figure 5. In the GPX10 Pro SoC, for instance, two clusters of five MX8 cores provide for high inference performance at ultra-low power for edge and endpoint applications. The GPX10 Pro can operate as a SoC, performing control and sensor interfacing operations as well as neural network processing, thanks to its Arm Cortex-M4F controller core and associated memory and peripherals; see Figure 6. The MX8 AI cores are easily
scaleable to meet the needs of different applications. The roadmap sees future GPX products scaling to 8,000 cores in the GPX8000 for data centre servers and supercomputers. And while the technology underlying
the GPX family of devices is new, developers can continue to work in a familiar workflow. The GPX family of processors is compatible with the main machine learning frameworks, including TensorFlow, PyTorch, Keras and ONNX. The Ambient Scientific software
development kit (SDK) for GPX devices includes a full model training toolchain. Once the application’s model has been trained, the developer uses Ambient Scientific’s Nebula SDK – a toolchain that enables complete AI application development. This includes a tool for compiling AI models to MX8 cores, as well as tools for configuring middleware, such as device drivers, real-time operating system and more, to run on the device’s Cortex-M4F core. Thus, developers can use familiar
Figure 6: The GPX10 Pro is not just an AI processor – it operates as a complete SoC
platform software for model development and achieve the same design productivity with the Ambient Scientific IDE as they are used to with conventional microcontrollers.
www.electronicsworld.co.uk February 2026 29
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48
