Neuromorphic Edge AI Chip Debuts on Raspberry Pi and Comet Lake Development Kits


BrainChip has launched two “Akida Development Kits” for its self-learning, power-saving neural network chip “Akida NSoC” for edge AI. One uses the Raspberry Pi CM4 and the other a shuttle PC system based on Comet Lake S.

BrainChip Holdings has opened pre-orders for two development kits showcasing its Akida Neural Networking Processor (Akida NSoC). The Linux-controlled Akida development kit for USD 4,995 – Raspberry Pi and Linux and Win 10 compatible Akida development kit for USD 9,995 – Shuttle PC implements the Akida NSoC via a mini PCIe module equipped with the AKD1000 silicon from BrainChip is.

Akida Development Kit – Raspberry Pi (left) and Akida Development Kit – Shuttle PC
(Click on images to enlarge)

The Akida NSoC is a neuromorphic, event-based AI processor that mimics brain processing, specifically the ability to “spike” processing. The Spiking Neural Networks (SNNs) enabled by the chip express information about both spatial and temporal sequences. As explained here EETimes report, spikes typically result from changes in sensor data, including color changes from an event-based camera.

In addition to SNNs, the Akida NSoC can also process standard convolutional neural networks (CNNs). The technology does this by converting CNNs to SNNs and performing inference in the event domain. This ability enables self-learning on the chip, which in turn enables the technology to handle changes in the perceived environment more flexibly than most AI chips. This self-learning and relearning ability can “eliminate the need for data round trips to centralized CPUs for retraining,” says BrainChip.

Akida IP architecture (left) and BrainChips benchmarks showing MAC operations required for object classification inference (dark blue is CNN in the non-event domain; light blue is Akida with the event domain; green is the event domain with further activity regularization)
(Click on images to enlarge)
Source: BrainChip via EETimes

Designed for edge computing, Akida NSoC reduces processing cycles and latency by focusing on key events while discarding “data of no value,” says BrainChip. This results in lower power consumption, with power budgets limited to microwatts or milliwatts, the company says. The technology is also touted for its performance, although we didn’t see a TOPS rating. EETimes has released the benchmarking chart for Akida MAC operations shown above.

The Akida NSoC architecture is based on an array of up to 20 nodes, each of which contains 4x neural processing units for a total of up to 80 NPUs. Like the brain, nodes are connected to one another via a mesh network. Scalability is further improved by node parallelization and the ability to use a smaller group of nodes “circumferentially”. Akida NSoC is powered by an “M-Class CPU with FPU and DSP”, which probably means a Cortex-M core.

Akida workflow diagram (left) and MetaTF development environment architecture
(Click on images to enlarge)

While many AI processors focus on vision and audio, Akida is also great for olfactory, gustatory, and vibration / tactile sensor applications. Applications include smart home, urban, transportation and health, with specific examples such as home automation and remote controls, industrial IoT, robotics, security cameras, sensors, unmanned aircraft, autonomous vehicles, medical instruments, object recognition, sound recognition, smell and taste recognition, gesture control and Cybersecurity.

The Akida Development Kits come with BrainChips MetaTF development environment, a machine learning framework that leverages Python alongside tools and libraries like NumPy. MetaTF includes an Akida Execution Engine chip simulator as well as data-to-event converters and a “model zoo” of pre-trained models.

Akida Development Kit – Raspberry Pi

The Raspberry Pi development kit is housed in a 200.6 x 105 x 37 mm housing. Inside, a carrier board is powered by the up to 1.5 GHz quad-core Cortex-A72 Raspberry Pi Compute Module 4. You can choose between CM4 models with 1 GB to 8 GB LPDDR4 and 8 GB to 32 GB eMMC. A CM4 with WiFi / Bluetooth is optional, and there seems to be an additional option on the carrier board for BT 5.0 with BLE.

Akida Development Kit – Raspberry Pi top view (left) and inside view with Akida-equipped mini PCIe card
(click on pictures to enlarge)

The Akida Development Kit – Raspberry Pi is equipped with a microSD slot together with GbE, 2x HDMI, 2x USB and a micro USB client port. A U.FL antenna connector and a fan are also available.

There are 2x 100-pin high-density connectors for connecting the CM4 as well as 40-pin GPIO, 2x MIPI-DSI and 2x MIPI-CSI interfaces. The specifications also include PCIe, but this is likely to be consumed by the mini PCIe slot with the Akida module.

Akida Development Kit – Shuttle PC

The Intel-based version of the kit is a modified version of a Shuttle PC computer equipped with a 10th generation Comet Lake-S processor. The 250 x 200 x 78.5 mm desktop PC overshadows the Shuttle PC mini PCs we have dealt with, such as the EN01 based on Apollo Lake.

Akida Development Kit – Rear view of the Shuttle PC (left) and interior view
(Click on images to enlarge)

The system supports up to 64 GB DDR4 and offers a 2.5-inch SATA bay and an M.2 M-Key 2280 slot with SATA and NVMe support. An M.2 E-Key 2230 slot supports an optional WiFi / BT module. There is also a full-size PCIe 3.0 x16 slot for up to 75 W cards.

The Akida Development Kit – Shuttle PC offers GbE, HDMI 2.0a and VGA connections. There are also 4x USB 3.2 Gen1 and 5x USB 2.0 ports, one of which is an internal USB 2.0 port for a USB stick. Further features are 2x audio jacks, an optional COM port, VESA bracket and a 180 W, 19.5 V adapter. Two fans with heat pipe support a range of 0 ~ 50 ° C.

additional Information

The Akida Development Kit – Raspberry Pi can be pre-ordered for $ 4,995 and the Akida Development Kit – Shuttle PC can be ordered for $ 9,995. No shipping date was specified. Further information can be found in BrainChips Notice and Akida Development Kits product page. You can find out more about Akida NSoC on the Akida NSoC page and that in detail User Guide.


Leave A Reply