Telecoms, Datacoms, Wireless, IoT


Sakura debuts for edge AI

22 November 2023 Telecoms, Datacoms, Wireless, IoT

EdgeCortix has made a strategic shift from selling AI intellectual property (IP) to selling its own edge-AI inference chips for line-powered systems. The new die, dubbed Sakura, started as a test chip, but the company says customer interest convinced it to offer the chip as a product. The chips come mounted on one of two cards: a dual-M.2 (M-key) card and a low-profile PCIe card. Based in Japan, EdgeCortix has supplemented its $13,5 million in total funding with revenue from FPGAs and ASICs that employ its soft IP. It targets perception – vision, lidar, and related technologies – are aimed at transportation, augmented/virtual reality, industry, smart cities, and drones.

Sakura, revealed first at the recent Linley Spring Processor Conference, implements the company’s dynamic neural accelerator (DNA) engine, adding on-chip SRAM, two LPDDR4X ports, and I/O. The chip has no host CPU, so it operates under the control of an external host. Sakura has a maximum performance of 40 TOPS; on ResNet-50, it achieves 0,4 ms latency at 4,7 W, yielding 533 inferences per second per watt (IPS/W).

Sakura combines six engines for standard 3D convolutions, six engines for 2D depth-wise convolutions, and three vector engines for activations and other miscellaneous operations. Standard convolutions combine all three colour channels, whereas depth-wise convolutions separate the channels, combining results later to reduce computing operations. To boost utilisation, the Mera tools can convert between standard and depth-wise convolutions to keep all engines occupied. Together, the engines are capable of 25 000 MACs per clock cycle.

Within a vector engine, dedicated blocks implement activation functions. Each block is configurable for a range of variants in a family of activation functions. Although this approach yields better performance, it’s less flexible than a fully programmable engine. A 20 MB on-chip SRAM can hold weights and activations or act as a scratchpad. The runtime software reconfigures the network-on-a-chip (NoC) on the fly, reassigning engines as part of EdgeCortix’s focus on achieving high hardware utilisation.

RFiber Solutions has added this group to its line card.


Credit(s)



Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

Simplifying system design
iCorp Technologies Telecoms, Datacoms, Wireless, IoT
The FGMC63N tri-band MCU Wi-Fi 6E module from Quectel addresses offers integrated connectivity across 2,4 GHz, 5 GHz, and 6 GHz bands in a compact, highly optimised design.

Read more...
Dragino’s connectivity catalogue
Otto Wireless Solutions Telecoms, Datacoms, Wireless, IoT
Where Wi-Fi demands frequent charging and covers tens of meters, and traditional cellular consumes too much power for battery-operated nodes, LoRaWAN threads the needle.

Read more...
RF filters explained
RFiber Solutions Telecoms, Datacoms, Wireless, IoT
As RF environments become increasingly crowded, RF filters play a critical role in maintaining signal quality, improving system performance, and ensuring reliable operation.

Read more...
Practical design strategies for 5G
RF Design Telecoms, Datacoms, Wireless, IoT
The Cavli CQM211 aims to provide a platform that combines strong 5G connectivity with onboard compute capability, as well as interface flexibility that suits both new designs and rapid migrations from earlier products.

Read more...
GigaDevice expands optical communication portfolio
NuVision Electronics Telecoms, Datacoms, Wireless, IoT
GigaDevice has introduced the new GD32E512 and GD32E252 series MCUs specifically designed for optical module applications.

Read more...
An industry-first Android 16 smart module solution
RF Design Computer/Embedded Technology Telecoms, Datacoms, Wireless, IoT
The MeiG Smart SLM580 4G Smart Module features native support for Android 16 and covers core frequency bands in major countries and regions worldwide.

Read more...
High performance RF signal conditioning
RFiber Solutions Telecoms, Datacoms, Wireless, IoT
The AM3269 from Mercury Systems is a high-performance RF preselector module designed to improve receiver sensitivity and signal integrity in demanding electronic warfare, signals intelligence, test, and communications applications.

Read more...
RF filters explained
RFiber Solutions Telecoms, Datacoms, Wireless, IoT
As RF environments become increasingly crowded, RF filters play a critical role in maintaining signal quality, improving system performance, and ensuring reliable operation.

Read more...
GigaDevice expands optical communication portfolio
NuVision Electronics Telecoms, Datacoms, Wireless, IoT
GigaDevice has introduced the new GD32E512 and GD32E252 series MCUs specifically designed for optical module applications.

Read more...
Industrial connectivity without compromise
Telecoms, Datacoms, Wireless, IoT
The IR315 combines 4G LTE, Wi-Fi, Ethernet, and VPN technologies into a rugged, industrial-grade platform that enables seamless connectivity across a wide range of applications.

Read more...









While every effort has been made to ensure the accuracy of the information contained herein, the publisher and its agents cannot be held responsible for any errors contained, or any loss incurred as a result. Articles published do not necessarily reflect the views of the publishers. The editor reserves the right to alter or cut copy. Articles submitted are deemed to have been cleared for publication. Advertisements and company contact details are published as provided by the advertiser. Technews Publishing (Pty) Ltd cannot be held responsible for the accuracy or veracity of supplied material.




© Technews Publishing (Pty) Ltd | All Rights Reserved