Yesterday, AMD unveiled its latest innovation: the Embedded+ architecture, sparking the question: why limit yourself to one compute architecture when you can have five?
This new offering from AMD merges a x64 Ryzen processor with a Versal AI Edge system-on-chip through PCIe. The objective? To offer a unified solution on a single board for low-power, low-latency data processing tasks, especially at the network edge.
At the core of this architecture lies the Ryzen Embedded R2000 family processor, introduced in 2022. It features up to four Zen+ CPU cores, 16 lanes of PCIe 3.0, and up to eight Radeon Vega graphics compute units.
This processor is directly connected via PCIe to an AMD Versal Adaptive SoC. These Versal chips, unveiled in 2021, come equipped with a variety of AI engines, an FPGA, and four Arm-designed CPU cores – two Cortex-A72 and two Cortex-R5. AMD claims that its top Versal chips can achieve around 228 TOPS at INT8 for ML processing.
Named Embedded+, this architecture is designed for use in devices operating in demanding environments such as public displays, field instrumentation, network edge processing, transportation, and automotive applications. While not necessarily cutting-edge in terms of performance, these chips prioritize factors like reliability, cost-effectiveness, power efficiency, compactness, and compatibility with specific workloads.
AMD has identified several industries as potential beneficiaries of this technology, including industrial robotics, retail, surveillance security, smart cities, networking, machine vision, and medical imaging. However, the suitability of the hardware for these applications will ultimately be determined by AMD's customers.
Chetan Khona, AMD's senior director of industrial vision, healthcare, and science markets, stressed the significance of low latency in automated systems and industrial and medical applications. In these scenarios, timely processing of sensor data is essential for enabling rapid and deterministic responses.
To meet the desired latency targets, AMD advises developers to break down their workloads into smaller tasks that can be individually accelerated by the platform's diverse compute architectures. For instance, the FPGA and AI engines within the Adaptive SoC can handle preprocessing and classifying streaming data from multiple sensors, while the CPU and GPU cores of the Ryzen processor manage control systems and graphical user interfaces.
While the concept of combining multiple architectures on a single board or chip is not new, AMD's approach distinguishes itself by integrating the Ryzen and Versal families, with a particular focus on AI capabilities at the embedded and network edge. This strategic move reflects AMD's responsiveness to market demands and its commitment to delivering solutions tailored to emerging application requirements.