BrainChip Unveils Ultra-Low Power Akida Pico Co-Processor for Next-Gen Portable Devices

October 3, 2024
BrainChip Unveils Ultra-Low Power Akida Pico Co-Processor for Next-Gen Portable Devices
  • BrainChip Holdings Ltd, based in Laguna Hills, California, launched the Akida Pico on October 1, 2024, introducing a low-power acceleration co-processor designed for compact, ultra-low power portable devices.

  • The Akida Pico targets a wide range of applications, including consumer electronics, healthcare, IoT, defense, and voice activation, making it versatile for various industries.

  • This co-processor is capable of executing specific neural network models with a power consumption of less than one milliwatt, making it particularly suitable for battery-operated devices.

  • Built on the Akida2 event-based computing platform, the Akida Pico operates efficiently under 1 milliwatt, enhancing its suitability for continuous monitoring systems.

  • The chip offers secure personalization features for applications such as voice wake detection, keyword spotting, and audio enhancement, enhancing user interaction.

  • Designed to wake up microcontrollers while minimizing power consumption, the processor filters out false alarms until an event is detected, contributing to energy savings.

  • BrainChip claims substantial power savings with the Akida Pico, potentially reducing energy consumption significantly compared to traditional models.

  • Key benefits of the Akida Pico include an ultra-low power neural processing unit (NPU) core, minimal standby power consumption, and a compact logic die area, optimizing it for energy efficiency.

  • CEO Sean Hehir emphasized that the Akida Pico is designed for users with varying levels of AI expertise, facilitating the creation of efficient neural networks.

  • The Akida platform is particularly ideal for low-latency applications, including robotics, drones, and automotive technologies, supporting early detection and responsiveness.

  • BrainChip's MetaTF software allows developers to optimize Temporal-Enabled Neural Networks (TENNs) for the Akida Pico using familiar frameworks like TensorFlow/Keras and Pytorch.

  • Despite the promising features, neuromorphic computing, including the Akida Pico, has yet to find widespread commercial applications, raising concerns about the capabilities of low-power AI solutions.

Summary based on 5 sources


Get a daily email with more Tech stories

More Stories