feature article
Subscribe Now

Xilinx Introduces Kria SoMs

Raising the Level of Integration

Integration is the fundamental fuel that drives technical innovation. Each new node of Moore’s Law has allowed us to “Cram More Components” onto our integrated circuits, facilitating greater and greater levels of integration on chunks of silicon. But even in today’s “System on Chip” era, we do not truly have “systems” on a chip.

Sure, we can pack processors, a lot of close peripherals, and some memory onto a single chip, but even our most integrated ICs pose challenges for bringing them into modern systems. The latest memory and data interfaces place heavy demands on board and module design for signal integrity and thermal performance, and the challenges of integrating the most sophisticated silicon devices we have today can be daunting.

One way to tackle those challenges is to have the semiconductor company pre-design modules with the most common interface challenges solved. This makes the task of bringing a sophisticated chip such as an SoC FPGA into our systems significantly simpler. And that is exactly what Xilinx has done with their newly-announced Kria SoM portfolio. Xilinx apparently figures you’ve got enough to worry about with your project without having to re-engineer interfaces that have been done a thousand times and that don’t really add differentiation to your design.

Kria SoMs are production-ready small form factor systems-on-modules that come with a pre-built software stack that should provide a great starting point for a wide variety of applications. The core idea is to allow teams with no or limited FPGA or hardware design experience to deploy systems without battling the usual obstacles that accompany design with advanced processors, accelerators, and interfaces. Xilinx claims that design teams can shave as much as nine months of development time from a typical project schedule by taking advantage of Kria.

Of course, an SOM with a built-in FPGA/SoC is interesting only if you have an application that needs the performance, flexibility, and power advantages of programmable logic. Today, many of those involve video analytics and processing – particularly in applications like security cameras, city cameras, traffic cameras, retail analytics, machine vision, and vision guided robotics.

Xilinx has built the Kria K26 around the Zynq UltraScale+ MPSoC, which includes a quad-core Arm Cortex A53 subsystem, over 250K logic cells of FPGA LUT fabric, and an H.264/265 video codec. The SoM also included 4GB 64-bit DDR4. This yields a theoretical 1.4 TOPS of AI compute performance. Kria boasts 245 I/Os via dual 240-pin connector, support for 15 cameras via a mix of MIPI, sub-LVDS, SLVS-EC, up to 40G Ethernet via 4x 10G interfaces, and 4x USB with a mixture of 2.0 and 3.0.

While the pre-built Kria relieves the design team of the burdens of developing memory interfaces and so forth, it also brings a lot of value on the software side. Most teams doing these types of applications have AI and software experts with little experience who want to get to a working implementation as fast as possible. The Kria SoM portfolio couples the hardware and software platform with production-ready vision-accelerated applications. These applications eliminate the FPGA hardware design work and allow software developers to easily integrate their custom AI models and application code. The Xilinx Vitis software development platform also gives developers the ability to easily modify the vision pipeline using design environments such as TensorFlow, Pytorch, or Caffé frameworks, as well as C, C++, OpenCL, and Python.

SoMs as platforms beg for third parties to develop turnkey applications and utilities that could be used by teams in a wide range of industries to shortcut their design process. In response to that notion, Xilinx has created an embedded app store for edge applications. This takes advantage of the infrastructure already in place for Xilinx’s Alveo data center apps, extending it to support Kria SoMs with apps from Xilinx and their partners. The apps developed by Xilinx are open-source accelerated applications, provided at no charge, including functions such as smart camera tracking, face detection, and natural language processing with smart vision.

As one would expect, Xilinx has development kits for Kria – the Kria KV260 Vision AI Starter Kit. Xilinx says the kit “provides an affordable and easy-to-use development platform for designing vision applications, right out of the box.” The kit supports accelerated vision applications from the Xilinx App Store, and the company claims it allows designers to “get up and running in less than an hour with no knowledge of FPGAs or FPGA tools.”

By making Kria a production-ready module, and by providing a very accessibly priced starter kit, Xilinx is effectively opening the technology up to hobbyists and makers as well as to commercial engineering teams. It will be interesting to see how the company handles the additional support burden that the former might introduce. It’s one thing to send a technical marketing engineer to a large commercial customer to help sort through design issues on a high-volume project. It is another thing entirely to field thousands of random and often ill-informed questions from hobbyists, including things like, “Can I power this with a 9V battery?” Xilinx says they will support Kria customers via “tutorial videos, training courses, and a vast ecosystem of providers offering accelerated applications, design services, and more.”

The Kria K26 SoM is available now (in a break from Xilinx’s tradition of long lead times on new product announcements), and targets vision AI in smart city and smart factory applications. The company is also planning a future cost-optimized version and a higher-end version aimed at maximizing real-time compute performance-per-watt for edge AI applications.

This notion of turnkey integration at the module level and app-store/open-source availability of reference designs and full-fledged applications signals Xilinx’s transition from a component company to a supplier of full-fledged solutions. This requires a completely new approach to customer engagement and support, which may be a challenge, given the company’s legacy. But making such a transition promises to open large markets for the company’s core technologies, and those markets would otherwise be largely inaccessible to them.

Leave a Reply

featured blogs
Jul 29, 2021
Circuit checks enable you to analyze typical design problems, such as high impedance nodes, leakage paths between power supplies, timing errors, power issues, connectivity problems, or extreme rise... [[ Click on the title to access the full blog on the Cadence Community sit...
Jul 29, 2021
Learn why SoC emulation is the next frontier for power system optimization, helping chip designers shift power verification left in the SoC design flow. The post Why Wait Days for Results? The Next Frontier for Power Verification appeared first on From Silicon To Software....
Jul 28, 2021
Here's a sticky problem. What if the entire Earth was instantaneously replaced with an equal volume of closely packed, but uncompressed blueberries?...
Jul 9, 2021
Do you have questions about using the Linux OS with FPGAs? Intel is holding another 'Ask an Expert' session and the topic is 'Using Linux with Intel® SoC FPGAs.' Come and ask our experts about the various Linux OS options available to use with the integrated Arm Cortex proc...

featured video

Design Success with Foundation IP & Fusion Compiler

Sponsored by Synopsys

When is 1+1 greater than 2? When using DesignWare Foundation IP & Fusion Compiler! Join Raymond and Yung in their discussion of a customer that benefited from the combination of Fusion Compiler’s machine learning and Foundation IP cells and macros.

More information about DesignWare Foundation IP: Embedded Memories, Logic Libraries, GPIO & PVT Sensors

featured paper

Configure the charge and discharge current separately in a reversible buck/boost regulator

Sponsored by Maxim Integrated

The design of a front-end converter can be made less complicated when minimal extra current overhead is required for charging the supercapacitor. This application note explains how to configure the reversible buck/boost converter to achieve a lighter impact on the system during the charging phase. Setting the charge current requirement to the minimum amount keeps the discharge current availability intact.

Click to read more

featured chalk talk

Power Profiler II

Sponsored by Mouser Electronics and Nordic Semiconductor

If you are working on a low-power IoT design, you are going to face power issues that can get quite complicated. Addressing these issues earlier in your design process can save you a lot of time, effort, and frustration. In this episode of Chalk Talk, Amelia Dalton chats with Kristian Sæther from Nordic Semiconductor about the details of the new Nordic Power Profiler Kit II - including how it can measure actual current, help you configure the right design settings, and show you a visualized power profile for your next design.

Click here for more information about the Nordic Semiconductor Power Profiler Kit II (PPK2)