Altera targets low-latency AI edge applications with new FPGA products

Altera has introduced the latest family of Agilex FPGAs, along with Quartus Prime Pro software, and FPGA AI Suite to enable the rapid development of highly customized embedded systems for use in robotics, factory automation systems, and medical equipment.

Altera, was acquired by Intel in 2015 but last year, Intel spun the FPGA maker out as a standalone business. The vendor has spent the better part of last year filling out things like accounting, human resources, and other general business operations.

The Agilex family of FPGA processors uses the same naming scheme as Intel’s Core brand the consumer use, with the brands 3, 5, 7, and 9, with 3 being on the low end of the spectrum, 9 on the top-of-the-line, and the other two in between.

This Embedded World conference announcement features the low-power, cost-optimized Agilex 3 FPGAs, which the vendor says delivers nearly double the fabric performance compared to the previous generation at up to 38% lower power. The FPGA’s lets businesses modernize their edge and embedded infrastructure by deploying customized AI solutions that deliver the low latency, energy efficiency and agility needed for system longevity, the company said.

“Having an AI infused fabric that allows you to configure that FPGA with the precise algorithms and capabilities and resources of the underlying platform to deliver on those AI tasks is really one of the benefits and appeals of an FPGA, that flexibility, that reprogrammability and the ability to run many different algorithms and customize the data paths that you need for your applications,” said Sandra Rivera, CEO of Altera in a conference call with journalists.

Agilex 3 FPGAs also support multi-arm robots with multi-axis arms by using machine learning capabilities. Agilex 3 FPGAs also support smart factory cameras to improve defect detection by using fine-grained parallel processing and Convolutional Neural Networks (CNNs) trained for object recognition to analyze vast amounts of data.

Support for Agilex 3 and other Agilex product lines is available through Altera’s free Quartus software suite. Quartus is a design software suite for programmable logic devices. It allows engineers to design, analyze, optimize, and program Intel FPGAs, CPLDs, and SoCs using system-level design techniques and advanced place-and-route algorithms.

For AI developers, Altera has upgraded its FPGA AI Suite to release 25.1, adding support for Agilex 3 and Agilex 5 FPGA development for AI inferencing using familiar industry-standard frameworks like TensorFlow and PyTorch along with OpenVino. “We’re just making it easier for our embedded customers to deploy more AI machine learning into their embedded platforms with FPGAs inside,” said Rivera.

In other Agilex news, the first wave of Agilex 5 FPGAs E-Series devices are now fully qualified and released for high-volume production. Compared to the Agilex 5 D-Series FPGAs, the Agilex 5 E-Series FPGAs are optimized for more power-sensitive applications that require high-performance with smaller form factors and logic densities.

Altera is also expanding the MAX 10 FPGA family with new package options. The MAX 10 10M40 and 10M50 product lines are now offered in variable pitch BGA packages. This new package option significantly increases the value of these highly integrated devices by reducing form factor while maintaining a high I/O count, resulting in a lower total cost of ownership for users the company claims.