All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Follow LinuxGizmos:
Twitter Facebook Pinterest RSS feed
*   get email updates   *

Neuropmorphic edge AI accelerator available in $499 mini-PCIe board

Jan 21, 2022 — by Eric Brown 582 views

BrainChip has launched a $499 “Akida AKD1000 Mini PCIe Board” equipped with its neuromorphic, event-based Akida neural networking chip for edge AI. The kit comes with BOM and design layout files.

In October, BrainChip Holdings launched two Akida Development Kits that combine a mini-PCIe board equipped with its Akida AI accelerator along with a Raspberry Pi 4 or Comet Lake-S based Shuttle PC. The company has now followed up with a version of the mini-PCIe implementation as a standalone product called the Akida AKD1000 Mini PCIe Board. To assist system integrators, BrainChip is providing the $499 board with full PCIe design layout files and the bill of materials (BOM) “to enable them to build their own boards and implement AKD1000 chips in volume as a stand-alone embedded accelerator or as a co-processor.”

Akida AKD1000 Mini PCIe Board (image on right appears to lack RAM block)
(click images to enlarge)

The 76 x 40 x 5.3mm, 15-gram Akida AKD1000 Mini PCIe Board is referred to as a board because it is larger than the standard full-sized 50.95 x 30mm full-sized mini-PCIe card spec. The product, which we saw on EENewsEmbedded, comes with a real-panel bracket (not shown).

The board incorporates a PCIe 2.0 x1 interface and “256M x 16 bytes LPDDR4 SDRAM @ 2400MT/s,” which we believe translates to 500MB. There is also 128Mb NOR QSPI flash, a core current monitor, and a pair of LEDs.


The AKD1000 chip, which was previously referred to as the Akida NSoC, is built around a 300MHz Cortex-M4 chip. The AKD1000 is a neuromorphic, event-based AI processor that mimics brain processing, especially in regard to the brain’s capability for “spiking” processing. The spiking neural networks (SNNs) enabled by the chip express information via both spatial and temporal sequences. Spikes typically result from changes in sensor data, including color changes from an event-based camera.

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

Although it is not designed to run standard convolutional neural networks (CNNs), Akida offers a mechanism to convert CNNs into SNNs so they can run inference in the event domain. This capability allows for on-chip self-learning, which in turn enables the technology to handle changes in the perceived environment with more flexibility than most AI chips.

Built for edge computing, Akida reduces processing cycles and latency by focusing on key events while discarding “data with no value,” says BrainChip. This translates into lower power consumption, with power budgets limited to microwatts or milliwatts, says the company. (For more on Akida, see our earlier Akida report.)

Akida Development Kit — Raspberry Pi (left) and Akida Development Kit — Shuttle PC
(click images to enlarge)

The Akida Development Kits both run Linux, but there was no mention of OS support for the standalone mini-PCIe offering. This report from 2020 says the SDK supports both Linux and Windows.

Further information

The Akida AKD1000 Mini PCIe Board is available for $499, with 8-week lead time. More information may be found in BrainChip’s announcement and product/shopping page.

(advertise here)

Print Friendly, PDF & Email

Please comment here...