Home>Posts>Tech Explained>What is AMD Embedded, and Why Now?

What is AMD Embedded, and Why Now?

By ·Categories: Tech Explained·Published On: January 20th, 2020·5.8 min read·

AMD Embedded is the newest offering in the industrial space. But what is AMD Embedded and why should you use it? Read on to find out.

AMD Embedded isn’t your typical consumer CPU hardware. While AMD Embedded processors have similarities to their consumer counterparts, AMD Embedded focuses on commercial and industrial deployments. That means, among other things, it delivers components with a long lifecycle and support—an important factor for industrial standardization. With offerings ranging from low power solutions, to high-end server products, AMD Embedded can be standardized across large-scale operations. This includes collections of edge devices, nodes or gateways, as well as the edge servers handling and controlling the entire infrastructure of a commercial or industrial enterprise.

AMD has made inroads into this market with their new Zen architecture, which introduce scalability across their product line. The AMD Zen architecture scales extremely well from low-end offerings starting at 6W all the way up to 225W powerhouse processors. The Zen architecture has been designed to increase the price to performance ratio, as well as performance per watt. This leaves end customers with a much lower Total Cost of Ownership (TCO) with affordable, reliable, and more efficient deployments. If you want to learn more about how watts impact processing power and cost, check out this article

Benefits of AMD Embedded

One unique AMD feature that allows for scalability is called Infinity Fabric architecture. When more compute performance is required, the Zen processor architecture can leverage Infinity Fabric to aggregate not only raw compute power, but also increase available I/Os. Each chiplet has 2 core complexes (CCX) of four cores, for a total of eight cores per chiplet, as well as a large cache bank. These chiplets all communicate with one another as well as the I/O controller die. The job of the I/O die is to handle all of the systems output interfaces, including PCIe, memory, and other sideband signals required. Instead of a single large silicon die, AMD has designed high-bandwidth interconnects allowing smaller dies to be created, which can be used across several product lines leading to increased yield and much lower manufacturing costs. This is only one area where AMD shines in a lower TCO for large scale Internet of Things (IoT) deployments.

A graphic of the AMD Ryzen Embedded logo

One of their products leveraging Infinity Fabric is a third-generation Ryzen “Matisse” processor.  This powerful offering supports up to two compute chiplets on one processor and an I/O die. AMD also scales that design up, creating high-end offerings with eight chiplets and an even larger I/O die to ensure their edge servers can handle a large number of devices efficiently. Furthermore, keeping the I/O die separate lets it take care of system memory, PCIe channels, and other signals while leaving the computing to the compute cores. This, in turn, leaves more physical space on the die to offer much more low level cache to further increase performance.

AMD Embedded scales in other ways to increase performance and lower costs. AMD’s new 7nm process offers higher transistor density, increasing performance and lowering power consumption. AMD also leverages the Zen architecture on 14nm and 12nm processes at a lower cost to find the right solution for each customer’s needs.

The excellent performance-per-watt and low cost of these processors make them an attractive choice for edge and IoT projects. Embedded computers need to deploy to a range of environments, including locations that are notoriously hostile to technology. Dust, debris, and corrosives created during production quickly destroy computers not designed for these environments. We address this issue using Hardshell Fanless Technology to seal systems from these contaminants and extend their longevity.However, fanless systems have a threshold of heat they can dissipate. This creates a trade off between efficiency, power, and size. AMD Embedded challenges those trade offs with capable low-watt processing. As a result, AMD mini PCs bring more horsepower to edge deployments without compromising size for compute.

Improved PCIe Speeds and Lanes

AMD’s new lineup adopts PCIe Gen 4.0, doubling the bandwidth of the previous PCIe Gen 3.0. As a result, PCIe solid-state NVMe drives that were bottle-necked by the PCIe bus are no longer constrained by these limitations. For customers requiring several high-bandwidth I/O cards, AMD offers up to 128 PCIe Gen 4.0 lanes in their Epyc™ line. Datacenter and collection nodes in particular benefit from the increased data processing and storage capacity. This allows customers with large deployments, which require several edge devices, to communicate quickly and efficiently to the AMD Epyc edge server.

Improved Cache Size and Performance

AMD has also been busy increasing the cache sizes of their processors. The 7nm process allows for increased transistor density, leading to a commensurate increase in potential cache capacity and lower power consumption. This on-die memory is where the processor stores the data, processing requests and outputs. Increasing the cache size increases the number of inputs and data it can process. This saves a lot of time in cache intensive applications, translating to snappy performance and quick task execution.

Improved Discrete Graphics Performance

CPU performance gains aren’t the only benefit to AMD Embedded technology. AMD also brings incredible graphics performance to the embedded space. When smaller industrial devices require GPU compute, solutions integrating AMD’s Radeon Vega graphics directly on the Silicon provides greater computational horsepower and support for more high-resolution displays than competitive offerings available on the market.

With Vega, AMD has realized two to three times the graphics performance of other integrated graphics as measured by FLOPs (Floating Point Operations per Second). This value defines how many 32-bit calculations the graphics processor can solve every second. This is crucial for the image-based processing tasks that are moving further out on the edge. Applications such as digital signage, medical imaging, casinos and entertainment benefit from this performance boost. These onboard discrete graphic processors address thermal concerns by offering excellent performance with low wattage. This allows graphics processing in smaller fanless systems relying on passive cooling to regulate temperature. The reliability of fanless computer systems results in lower mean time between failure.

Getting started with AMD Embedded computers

A photo of three OnLogic embedded industrial computers

With excellent cost-to-performance, optimized efficiency, and capable onboard graphics, AMD’s Embedded line of processors brings viable new options to the embedded market. Browse our line of AMD Embedded PCs spanning a range of form factors and configuration options to find one that meets your needs. There are a staggering number of options for embedded edge computers. We’ve created guides on how to find the right edge computer and the differences between edge computers to help your search. Contact the experts at OnLogic if you have more questions and don’t forget to subscribe to the I/O Hub to stay up to date on the latest AMD Embedded technology. 

Get the Latest Tech Updates

Subscribe to our newsletters to get updates from OnLogic delivered straight to your inbox. News and insights from our team of experts are just a click away. Hit the button to head to our subscription page.


About the Author: Alejandro Vinals

Alejandro is an electrical engineer on the OnLogic engineering team.