California-headquartered Xilinx ventured into the growing system-on-module (SOM) market with the launch of its Kria portfolio. This announcement comes when the SOM market is growing steadily and its total market revenue is expected to touch $2.5 billion by 2026.
SOM is an electronic circuit that integrates the required system functions into a single module that includes a processor, memory, and I/O controller.
“The Kria SOM portfolio expands our market reach into more edge applications and will make the power of adaptable hardware accessible to millions of software and AI developers,” said Kirk Saban, VP, Product and Platform Marketing at Xilinx. The company plans to further expand to the family of reconfigurable FPGA-based single board computers coupled to Arm core CPU engines.
With the launch of Kria K26 SOM, the company hopes to accelerate the use of AI, machine learning, and vision at the edge. Xilinx also launched the KV260 Vision AI starter kit, which provides an easy-to-use development platform for designing vision applications.
The K26 SOM is based on Zynq’s UltraScale + MPSoC architecture, which includes a quad-core ARM Cortex A53 processor, over 250,000 logic cells, and an H.264/265 video codec. The SOM also has 4GB of DDR4 memory and 245 I/O paths, allowing it to adapt to almost any sensor or interface. When compared to GPU-based SOMs, it enables developers to build AI applications with more than three times the performance at lower latency and power.
As per the company, Kria SOMs utilise the power, performance and flexibility of Xilinx adaptable hardware, which is delivered as adaptive modules that can be deployed in production. It offers end-to-end board-level solutions and a pre-built software stack. Developers get a headstart as Kria SOMs reduce the time-to-deployment by up to nine months.
Kria’s SOM portfolio couples the hardware and software platform with production-ready vision accelerated applications that eliminate the need for designing FPGA hardware from scratch. Software developers would only need to integrate custom AI models and application code to modify the vision pipeline. K26 offers design platforms such as TensorFlow, Pytorch or Caffe frameworks, C, C++, OpenCL™, and Python programming languages—enabled by the Vitis™ unified software development platform and libraries.
Embedded developers can now customise and optimise for standard Yocto-based PetaLinux. For the first time, Xilinx will also be offering Ubuntu Linux support for interoperability with existing applications.
Along with Kria K26 SOM, Xilinx also made another major announcement–Kria KV260 Vision AI Starter Kit availability. Customers can use the Kria KV260 Vision AI Starter Kit to test their target applications and, eventually, build their own carrier card using Xilinx K26 SOMs. Smart city and machine vision, surveillance cameras, retail analytics, and other industrial applications are among the Kria KV260 Vision AI Starter Kit applications.
Kria KV260 Vision AI Starter Kit (Credit: Xilinx)
Xilinx has also announced an embedded app store for edge applications that offers customers a selection of Kria SOMs from Xilinx and other partners.
It offers open source accelerated applications in facial recognition to natural language processing.
Growing SOM market
SOMs offer high-performance, low-power requirements and good connectivity. These features make them ideal for mobile devices, smart homes, and other IoT devices.
Further, the faster development in the automotive industry is also anticipated to drive the SOM demand. Other factors accelerating SOM demand include the need for uninterrupted communication and wireless connectivity.
As per the Global Market Insights 2020 report, the industrial automation application segment held the mammoth SOM market share at 45 percent in 2019. It is expected to expand at 12 percent CAGR till 2026, driven by the development of industrial robots in the manufacturing sector. It is followed by industries such as medicine, entertainment, transportation, testing, and measurement.