AMD has announced the expansion of the AMD Versal adaptive system on chip (SoC) portfolio with the new Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 adaptive SoCs, which bring preprocessing, AI inference, and postprocessing together in a single device for end-to-end acceleration of AI-driven embedded systems.
The initial devices in the Versal Series Gen 2 portfolio build on the first generation with powerful new AI Engines expected to deliver up to 3x higher TOPs-per-watt than first generation Versal AI Edge Series devices, while new high-performance integrated Arm CPUs are expected to offer up to 10x more scalar compute than first gen Versal AI Edge and Prime series devices.
“The demand for AI-enabled embedded applications is exploding and driving the need for single-chip solutions for the most efficient end-to-end acceleration within the power and area constraints of embedded systems,” says Salil Raje, senior vice-president and GM: Adaptive and Embedded Computing Group at AMD.
“Backed by over 40 years of adaptive computing leadership, these latest generation Versal devices bring together multiple compute engines on a single architecture offering high compute efficiency and performance with scalability from the low-end to high-end.”
Versal Series Gen 2 devices enable the design of high-performance, edge-optimised products for the automotive, aerospace and defense, industrial, vision, healthcare, broadcast and pro AV markets.
To meet the complex processing needs of real-world systems, AMD Versal AI Edge Series Gen 2 devices incorporate an optimal mix of processors for all three phases of AI-driven embedded system acceleration:
* Preprocessing: FPGA programmable logic for real-time preprocessing with unparalleled flexibility to connect to a wide range of sensors and implement high-throughput, low-latency data-processing pipelines
* AI Inference: An array of vector processors in the form of next-gen AI Engines for efficient AI inference
* Postprocessing: Arm CPU cores providing the postprocessing power needed for complex decision-making and control for safety-critical applications.
This single-chip intelligence can eliminate the need to build multi-chip processing solutions, resulting in smaller, more efficient embedded AI systems with the potential for shorter time-to-market.
The AMD Versal Prime Series Gen 2 provides end-to-end acceleration for traditional, non-AI-based embedded systems by combining programmable logic for sensor processing with high-performance embedded Arm CPUs. Designed to offer up to 10x more scalar compute compared to the first generation, these devices can efficiently handle sensor processing and complex scalar workloads.
With new hard IP for high-throughput video processing, including up to 8K multi-channel workflows, Versal Prime Gen 2 devices are ideally suited for applications such as ultra-high-definition (UHD) video streaming and recording, industrial PCs, and flight computers.
The Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 portfolios provide scalability from edge sensors to centralized compute for AI-driven systems. They feature a range of devices with increasing amounts of AI and adaptive compute to allow customers to select the performance, power, and area footprints to efficiently achieve application performance and safety targets.
The AMD Vivado Design Suite tools and libraries help boost productivity and streamline design cycles for embedded hardware system developers, offering fast compile times and enhanced quality of results. For embedded software developers, the AMD Vitis Unified Software Platform enables embedded software, signal processing, and AI design development at users’ preferred levels of abstraction, with no FPGA experience needed.
Designers can get started with AMD Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 early access documentation and first-generation Versal evaluation kits and design tools available today.
AMD expects availability of Versal Series Gen 2 silicon samples in the first half of 2025, followed by evaluation kits and System-on-Modules samples in mid-2025, and production silicon expected in late 2025.