The Akida Neuromorphic IP is the first neuromorphic IP available in the market. Inspired by the biological function of neurons but engineered on a digital logic process, this event-based spiking neural network (SNN) IP is inherently lower power than traditional convolutional neural networks (CNN) accelerator IP. When using the unique BrainChip CNN2SNN conversion flow, the event-based nature of SNNs enable converted CNNs to be implemented with very low power consumption and high throughput. Because the IP is based upon neuromorphic SNN, it enables unsupervised learning allowing for autonomous edge applications. The Akida Neuromorphic IP contains interfaces through a standard AXI bus making it easily integrated with any ASIC controller.
BrainChip’s AI IP is an event-based technology that is inherently lower power when compared to conventional neural network accelerators. BrainChip IP allows incremental learning and high-speed inferencing in a wide variety of use cases with high throughput and unsurpassed performance-per-watt.
BrainChip’s IP can be configured to perform convolutions (CNP) and fully connect (FNP) layers. Weight bit-precision is programmable to optimize throughput or accuracy and each weight is stored locally in embedded SRAM inside each NPU. The entire neural networks can be placed into the fabric, removing the need to swap weights in and out of DRAM resulting in a reduction of power consumption while increasing throughput.
BrainChip’s IP fabric can be placed either in a parallelized manner that would be ideal for ultimate performance, or space-optimized in order to reduce silicon utilization and further reduce power consumption. Additionally, users can modify clock frequency to optimize performance and power consumption further.
- Robust software and development environment and tools
- Complete configurable neural network processor
- On-chip mesh network interconnect
- Standard AXI 4.0 interface for on-chip communication
- Scalable nodes can be configured as:
- – Event domain convolution neural processor
- – Fully connected neural processor
- Hardware-based event processing
- No CPU required
- External memory optional (SRAM or DDR)
- Configurable amounts of embedded memory and input buffers
- Integrated DMA and data-to-event converter
- Hardware support for on-chip learning
- Hardware support for 1b, 2b or 4b hybrid quantized weights and activations to reduce power and minimize memory footprint
- Fully synthesizable RTL
- IP deliverables package with standard EDA tools
- – Complete testbench with simulation results
- – RTL synthesis scripts and timing constraints
- – Customized IP package targeted for your Application
- 4X power and memory savings over state of the art NN accelerators
- Complete Neural Network Processor completely offloads CPU
- Minimize integrated or external memory footprint
- Complete object detection and classification neural network solutions available utilizing MobileNet and MobileNet SSD
- Run muttiple networks in parallel or re-configure on the fly in milli-seconds
- Use Cases:
- Object/Face detection and classification
- Object tracking
- Sound detection and classification
- Keyword spotting
- Gesture recognition
- Cybersecurity Packet inspection
- The Akida Neuromorhpic IP is available as standard RTL for designing into embedded systems.
- Additionally, BrainChip will provide testbenches and other standard IP integration support infrastructure.
- Smart appliances
- Remote controls
- Industrial IoT
- Security cameras
- Automotive - ADAS/AV
- Audio devices
Block Diagram of the Complete Neural Processor for Edge AI.