Intel announces edge AI processors

Intel and Altera, an Intel company, announced new edge-optimized processors that are specifically aimed at major vertical industries such as retail, manufacturing, healthcare, and aerospace. Unveiled at the Embedded World conference, the lineup consists of FPGA processors from Altera as well as Atom and Core CPUs and Arc discrete graphics.

One of the advantages of Intel’s edge AI processors is they are integrated into other Intel products and tied together with Intel’s OneAPI library.

“This next generation of Intel edge-optimized processors and discrete GPUs unleashes powerful AI capabilities to help businesses more seamlessly incorporate AI alongside compute, media, and graphics workloads,” said Dan Rodriguez, Intel’s corporate vice president and general manager of network and edge solutions group, in a statement.

Intel said the Agilex 5 SoC FPGAs are now broadly available to customers. The Agilex 5 (pictured above) is aimed at mid-level devices, and Intel claims up to twice the performance per watt versus competitive 7 nanometer FPGAs. They are designed to deliver high performance with lower power in a modern SoC subsystem with small form factor package options.

Agilex 5 development kits are also broadly available. The FPGA AI Suite allows developers to use existing and popular AI frameworks, along with Intel’s OpenVINO toolkit, to create AI intellectual property blocks and easily drop them into the FPGA design.

Intel also updated its Quartus Prime Design Software suite for building the FPGA-based apps. The latest version of Altera’s development software adds support for the Agilex 5 E-Series SoC FPGAs and selected complementary IP cores.

In non-FPGA news, Intel announced Intel Core Ultra Processors for Edge. These processors come with Intel Arc GPU and a neural processing units (NPU) and offer up to five times better image classification inference performance compared to previous generations.

The processors are designed for applications in retail, education, smart cities, and industrial sectors, supporting generative AI and graphics workloads at the edge. Intel also has lower power Core processors for edge as well. These processors offer up to 2.57 times greater graphics performance compared to previous generations.

Intel’s Atom processors, its failed attempt to compete with Arm in the embedded space, lives on in the x7000C series tailored for enterprise networking and telecommunications devices. These processors offer packet processing and deep learning inference capabilities functionality.

The Atom x7000RE Series is designed for industrial and manufacturing applications with built-in deep learning inference capabilities and up to 32 graphics execution units. They support fanless designs and enable AI-automated tending, warehouse automation, and quality control applications.