AMD Collaborates with Arm to Unveil Next-Gen AI Chip Family

In a move poised to reshape the landscape of AI-driven embedded systems, AMD has joined forces with Arm to introduce a groundbreaking chip family, promising end-to-end acceleration for a variety of applications. The collaboration aims to address the burgeoning demand for efficient AI solutions across diverse sectors, from autonomous driving to healthcare.

The newly announced Versal AI Edge Series Gen 2 adaptive SoCs (System on Chips) represent a significant leap forward in AI processing technology. Unlike traditional multi-chip solutions burdened with overheads, AMD’s innovative approach integrates preprocessing, AI inference, and postprocessing onto a single silicon platform. This streamlined architecture not only enhances performance but also optimizes power and area constraints inherent in embedded systems.

At the heart of the Versal AI Edge Series Gen 2 SoCs lies a sophisticated processing system, blending Arm CPUs with next-generation AI Engines and programmable logic. This fusion of computational prowess enables seamless handling of the diverse computational phases crucial for embedded AI applications.

Designed with a focus on safety, security, and reliability, these SoCs are primed for a wide array of markets, ranging from automotive to healthcare. With features such as up to 8x Arm Cortex-A78AE application processors and support for various interfaces like USB 3.2 and PCIe Gen5, the Versal AI Edge Series Gen 2 SoCs offer unparalleled versatility and performance.

Salil Raje, Senior Vice President of AMD’s Adaptive and Embedded Computing Group, emphasized the significance of this collaboration in meeting the escalating demand for AI-enabled embedded applications. He underscored the efficiency and scalability of the Versal devices, which draw on AMD’s four-decade legacy in adaptive computing.

While the announcement heralds a new era in AI processing, eager customers will need to exercise patience. Initial silicon samples of the Versal Series Gen 2 are slated for release next year, with production expected to commence in late 2025. In the interim, early access documentation and evaluation kits are available, offering a glimpse into the future of embedded AI computing.