The tape-out was completed using GlobalFoundries’ 22nm fully depleted silicon-on-insulator (FD-SOI) technology and is being described as a milestone in validating BrainChip’s IP across different processes and foundries, providing its partners with varied global manufacturing options.
The AKD1500 uses the latest features of BrainChip’s energy-efficient, event-based, Akida AI IP in combination with the GlobalFoundries’ ultra-low leakage, proprietary FDX process technology platform, providing always-on, at-sensor applications or other low-power embedded applications for the AIoT, industrial, consumer and automotive markets.
In always-on, at-sensor use cases, Akida can completely offload feed-forward networks without the use of a CPU, ensuring energy efficient intelligent sensors. For other low-power embedded applications, the CPU solely performs sensor fusion processes or runs the application with Akida managing all the AI acceleration to significantly minimise CPU workload and system load, freeing up the MCU for other general-purpose computation.
The AKD1500 has been designed as an accelerator reference chip, using quad/octal SPI interfaces for MCU modules and PCIe for extensible cards, which assists partners in developing and demonstrating their solutions as a steppingstone to integrating the Akida IP into their production SoCs.
“This is an important validation milestone in the continuing innovation of our event-based, neuromorphic Akida IP, even if it is fully digital and portable across foundries,” said Anil Mankar, Co-Founder & Chief Development Officer at BrainChip. “The AKD1500 reference chip using GlobalFoundries’ very low-leakage FD SOI platform, showcases the possibilities for intelligent sensors in edge AI.”
