Xilinx has entered the System on Module (SOM) markets with the Xilinx Kria SOM and Vision AI starter kit. Xilinx Kria SOM K26 is a new way of deploying Xilinx platforms. It comes along with pre-built hardware, software that are production-ready for accelerated and AI applications. Being flexible to changing AI and sensor requirements, it also promises to be adaptable to dynamic AI requirements while working at low power and low latency. Thus, Xilinx Kria brings AI acceleration to the edge.
The Kria K26 SOM features Zynq UltraScale+ MPSoC that provides a processing system with integrated programmable logic. Being based on an architecture that possesses a quad-core Arm Cortex-A53 processor with 256K system logic cells, 4k 60p H.264/265 video codec and 1.4TOPS AI processing performance. Kria SOM has raised bars in SOM markets. Moreover, the board comes with a 4GB of 64-bit DDR4 memory, about 245 general-purpose IOs, 15 camera supporting ports, 4 USB ports (2.0, 3.0) and a 40 Gbps compatible ethernet port. All of these peripherals make Kria K26 SOM ideal for AI machine vision and inference applications.
Inclusions and Kria portfolio
The starter kit includes K26 SOM, a carrier card and the thermal solution. Besides this credit card-sized board comes in commercial and industrial temperature grades. The small size is a perfect fit for production deployment in Smart Camera, Embedded Vision, Security, Retail Analytics, Smart City, and Machine Vision applications. Kirk Saban, the vice president of Product and Platform Marketing at Xilinx, confirmed the lined up portfolio of Kria SOM. He stated, “Xilinx Kria will eventually expand into a family of single-board computers based on reconfigurable FPGA (Field Programmable Gate Array) technology, coupled to Arm core CPU engines and a full software stack with an app store, the first of which is specifically is targeted at AI machine vision and inference applications.”
Xilinx supports two ways for developers to engage with SOM. Firstly, the SOM mounted on Xilinx built carrier card evaluation and deployment. Another is the case where developers use user-built carrier card production and deployment. “The KV260 Vision AI starter kit target AI applications. It is enabled by several accelerated applications.” In addition, the board enables users to target the SOM on the starter kit as no prerequisite of installing and FPGA knowledge is required. Its design is specifically for users to apply their target applications, followed by developing their carrier card (K26 SOM) as per the application requirements. Hence, Xilinx Kria brings AI acceleration to the edge.
Xilinx Kria and AI Acceleration
The Xilinx Kria SOM aims at the AI audience. Besides, it is compatible with AI native frames that include Tensor flow, Pytorch Café frameworks, C, C++, and OpenCL. Thus, a developer can modify vision pipelines as per the need. “The AI developers can replace the AI model with a custom model and application code to make kit deployment ready. Software developers can start quickly with SOM and decouple parallelise their work without waiting for the first in-house prototypes. The Hardware developers can spend more time on what will differentiate their end systems as the SOM often takes care of lower value design work by using pre-built common functionalities like DDR and PS peripherals.”
Apart from this, the Xilinx Kria SOM will be collaborating with Ubuntu Linux. “Together with Xilinx, we’re excited to provide Kria SOM customers with out-of-the-box productivity, the frictionless transition from development to production, and guaranteed stability and security in the field,” said Thibaut Rouffineau, vice president of marketing at Canonical/Ubuntu. It will introduce Xilinx Kria to a bigger audience. Xilinx also announced the world’s first embedded app store for edge AI applications. Moreover, being free of charge this app store will be provided for accelerated open-source applications.
Sanskriti Sawant is a student of Electronics and Telecommunication engineering. She is passionate about Computer architecture, VLSI and plans to major in them. She is working on HDL languages and FPGA’s.