Highlights:

  • Organizations can customize Intel’s FPGA design to enhance performance speed for specific workloads or minimize power consumption.
  • Intel asserts that the product series, built on a 10-nanometer process, delivers up to 60% greater performance per watt than competitor products manufactured using seven-nanometer technology.

Intel Corp. recently introduced its independent FPGA business, offering reconfigurable chips tailored for various systems, from cell towers to robotics.

The origins of the business can be traced back to a chipmaker named Altera Corp., which was established in 1983. In 2015, Intel acquired the company for USD 16.7 billion and subsequently integrated it into its data center division. The previous year, it unveiled intentions to separate Altera from the data center division and establish it as an independent entity.

The transistors within a processor are structured into modules, each designed to execute a distinct set of functions. A particular module may be fine-tuned to execute matrix multiplications, the mathematical computations commonly employed by machine learning models to analyze data. In contrast, the adjacent cluster of transistors may specialize in encrypting network traffic.

Most processors do not offer users the ability to modify which cluster of transistors handles specific tasks or their configurations. Intel’s recently independent Altera unit offers specialized chips known as FPGAs, or field programmable gate arrays, which support customization. Organizations can customize Intel’s FPGA design to enhance performance speed for specific workloads or minimize power consumption.

For instance, a hardware team developing an AI-powered smart home appliance could optimize most circuits within an FPGA to execute matrix multiplications. The greater the number of AI-optimized circuits within a processor, the more efficiently it can execute machine learning models faster. FPGAs offer the capability to customize their onboard compute modules and the interconnects responsible for managing the data flow between these modules.

Altera is spearheaded by Chief Executive Officer Sandra Rivera, who formerly headed the Intel data center division from which the business was spun. During a recent webcast, Rivera stated that the business is poised to tap into an addressable market of USD 55 billion within the next few years. She also offered insights into Altera’s product development roadmap during the webcast.

Rivera offered a sneak peek into Agilex 3, an upcoming series of Intel’s FPGAs designed for low-power devices with specific computing capacity requirements. Altera envisions these chips being utilized in cloud, communications, and edge computing environments. The Intel unit plans to unveil more details about the product series later this year.

Soon, Altera will introduce a midrange family of Intel’s FPGAs called the Agilex 5 line. Intel asserts that the product series, built on a 10-nanometer process, delivers up to 60% greater performance per watt than competitor products manufactured using seven-nanometer technology.

Another notable feature is that the Agilex 5 has a digital signal processor, or DSP, which leverages AI-optimized circuits to accelerate computations. A DSP is a dedicated processor designed for tasks like compressing multimedia files and correcting any transmission errors they may encounter. Other computation modes are also included in the Agilex 5, such as central processor unit cores based on an Arm Holdings plc design.

Intel also recently shared an update on Altera’s previously unveiled Agilex 7 and Agilex 9 product lines. The company announces that the Agilex 7 chip family has been released to production, whereas the Agilex 9 family is currently in volume production.

The Agilex 7 is optimized for deploying data centers, networks, and defense systems.
Conversely, the Agilex 9 is tailored for radar and military aerospace applications. Both product lines have been optimized to handle mixed-signal processing, which involves the simultaneous processing of standard data encoded as ones and zeros and other signal types, including radar readings.

Rivera stated, “As customers deal with increasingly complex technological challenges and work to differentiate themselves from their competitors and accelerate time to value, we have an opportunity to reinvigorate the FPGA market.”

Altera’s primary competitor is Advanced Micro Devices Inc., which ventured into the FPGA market following its USD 50 billion acquisition of Xilinx in 2022. The acquisition represented the largest deal in the semiconductor industry’s history. According to International Data Corp., AMD commands a 55% share of the FPGA market following the Xilinx acquisition, while Altera holds over 30% of sales.