The MAX78000 reduces energy consumption and latency by 100 times, enabling complex embedded decisions at the IoT edge

Beijing, China—October 21, 2020—Maxim Integrated Products, Inc (NASDAQ: MXIM) announced the launch of a neural network acceleratorMAX78000 Low-power microcontrollers that support battery-powered embedded Internet of Things (IoT) devices at the edge through fast, low-power artificial intelligence (AI) reasoning to make complex decisions. Compared with software solutions, this fast and low-power decision-making implementation reduces the energy consumption of complex AI reasoning to less than one percent of the previous solution. The battery-powered system using AI technology can greatly extend its running time and help To achieve a new generation of battery-powered AI applications that cannot be surpassed before. In addition, the MAX78000 does not affect latency indicators and costs: its cost is only a fraction of the FPGA or GPU solution, and the speed of inference is 100 times faster than the software solution implemented on a low-power microcontroller.

• For more information about Maxim Integrated artificial intelligence solutions, please visit ›
• To order MAX78000 or for more information, please visit ›
• To download high-resolution images, please visit ›
• For more information about MAX78000 development tools and a quick start guide video, please visit ›

AI technology enables machines to observe, listen and perceive the world in ways that were completely impossible before. In the past, placing AI inference to the edge meant collecting data from sensors, cameras, and microphones, then sending the data to the cloud to implement inference algorithms, and then sending the results back to the edge. Due to the large delay and energy consumption, this architecture is extremely challenging for edge popularization. As an alternative, low-power microcontrollers can be used to implement simple neural network operations, but the delay will be affected and simple tasks can only be performed at the edge.

By integrating a dedicated neural network accelerator, the MAX78000 overcomes these limitations, and enables the machine to see and hear complex patterns by executing AI processing locally with low power consumption in real time. Because the MAX78000 uses less than one percent of the power consumption of the microcontroller software to execute inference, it greatly improves the efficiency of applications such as machine vision, voice and facial recognition. The core of MAX78000 is dedicated hardware, which is designed to minimize the energy consumption and delay of Convolutional Neural Network (CNN). When the hardware is running, almost no microcontroller core is required to intervene, which means that the operation is extremely streamlined. Energy and time are only used to implement CNN mathematical operations. In order to efficiently input collected data from the outside world into the CNN engine, users can use one of two integrated microcontroller cores: ultra-low power ARM Cortex-M4 core, or lower power RISC-V core.

In view of the challenges of AI development, Maxim Integrated provides tools to achieve a smooth evaluation and development experience. MAX78000EVKIT# Including audio and camera input, the out-of-the-box presentation platform supports keyword retrieval and facial recognition in large print. Complete documentation can help engineers train the MAX78000 network, and use their daily tools: TensorFlow or PyTorch.

Main advantage

• Low energy consumption: The hardware accelerator is combined with ultra-low power consumption ARM M4F and RISC-V microcontrollers to push the intelligent implementation to the edge, and the energy consumption is less than one percent of the embedded competing solutions.
• Low latency: Perform AI functions at the edge to achieve complex cognition, reduce or eliminate cloud transaction processing for IoT applications, and increase the speed to 100 times that of software solutions.
• Highly integrated: Low-power microcontrollers with neural network accelerators make it possible to realize complex, real-time cognition in battery-powered IoT devices.

Evaluation

• "Artificial intelligence is often associated with big data cloud solutions." Omdia's market research analyst Kelson Astley said: "Any technology that can be separated from power lines and reliance on large-capacity lithium-ion battery packs will help open up designers' thinking. , And build an AI solution that is more agile and more adaptable to its operating environment."
• "We have eliminated the power cord for edge AI implementation." Kris Ardis, Executive Director of Maxim Integrated Microprocessor and Security Products Division, said: "Battery-powered IoT devices can now do much more than simple keyword recognition. We have It has changed the rules of the game that have to choose between power consumption, latency and cost. We look forward to this innovative technology to give birth to new application areas."