Intel's Altera Unit Unveils Ambitious Plans for Next-Gen Programmable Chips

Altera brings more AI to the edge and cloud with new programmable chips

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Altera, a division of Intel that makes programmable chips, unveiled today a number of products that will bring more AI to the edge and the cloud. The products include field programmable gate array (FPGA, or chips that can be programmed after being designed into products) hardware, software, and development tools that make its programmable solutions more accessible across a broader range of use cases and markets. At its annual developer's conference, Altera revealed new details on its next-generation, power- and cost-optimized Agilex 3 FPGAs and announced new development kits and software support for its Agilex 5 FPGAs, said Sandra Rivera, CEO of Altera, in a press briefing. "Working closely with our ecosystem and distribution partners, Altera remains committed to delivering FPGA-based solutions that empower innovators with leading-edge programmable technologies that are easy to design and deploy," Rivera said. "With these key announcements, we continue to execute on our vision of shaping the future by using programmable logic to help customers unlock greater value across a broad range of use cases within the data center, aerospace and defense sectors, communications infrastructure, automotive, industrial, test, medical and embedded markets." Rivera has been focused on Altera for about nine months, and today the company is holding its Altera Innovator's Day. She said the event represents the relaunching of the company under the Altera brand, largely separated from Intel. "Our goal is to be the No. 1 FPGA provider in the world. It's a big, audacious, ambitious goal for us. We're the only company left in the world that is top to bottom, cloud to edge FPGAs. "This is all we do every day, breathe, eat, sleep and drink the FPGA portfolio," she said. Why it matters Altera is not a small startup. It's got 40 years of history and was acquired by Intel in 2015 for $16.5 billion. Altera's arch competitor is Xilinx, another FPGA firm that was acquired in 2020 by Intel's rival AMD for $35 billion. At first, Intel wrapped Altera into its own operations. Earlier this year, Rivera began undoing that. This meant that Altera disengaged its operations, marketing, support, product and other teams from Intel. "From an overall product execution perspective, we made decisions early in the year to really refactor the roadmap, simplify things, lean into that waterfall strategy where we have more IP reuse, more engineering leverage for the investments, and, frankly, simplify not just our own product development, but that of our customers as well," Rivera said. "And that's been really well received." She added, "Our focus then is really entirely on how do we continue to build our pipeline and convert more of that pipeline to design wins. We feel really good about where we stand today in terms of just the design wins, and winning more than 50% of the of the opportunities that we go after. And this is how we get to number one." Rivera said Altera is the only independent FPGA supplier with full-stack solutions that are optimized across high-performance accelerated computing systems, next-generation communications infrastructure and intelligent edge applications. "We are just absolutely focused on ensuring that we are delivering best in class capabilities to our customers so that they can deploy at scale, easily and cost effectively, as well as just accelerating their time to value for deploying FPGA solutions as part of their overall compute platform," Rivera said. "There is a lot of excitement from our customers and from our ecosystem partners around that positioning and around the opportunities for us to really unlock more value and capability with a company that is solely dedicated to the FPGA industry, and that has a full portfolio of capabilities, so lots of great stuff is going on." The company's comprehensive FPGA portfolio provides customers with flexible hardware that rapidly adapts to changing market requirements driven by the era of intelligent compute, she said. Altera is targeting FPGAs for AI inference workloads with Agilex FPGAs infused with AI Tensor Blocks and the Altera FPGA AI Suite, which accelerates FPGA development for AI inference using popular frameworks such as TensorFlow, PyTorch, and OpenVINO toolkit and proven FPGA development flows. What Agilex 3 FPGAs offer Noting the company's introductions of its Agilex chips and pointing to the importance of AI, Rivera said, "We have the portfolio breadth, the performance, the software and, importantly, the AI capabilities that so many of our customers want on our path to No. 1." She noted the chips included the only product in the FPGA world with that AI infused fabric. Altera announced today new product details for its Agilex 3 FPGAs, designed to meet the power, performance, and size requirements of embedded and intelligent edge applications. Compared to the previous generation, Agilex 3 FPGAs bring higher levels of integration, enhanced security, and higher performance in a compact package, with densities ranging from 25,000 to135,000 logic elements. The FPGA family features an on-chip dual Cortex A55 ARM hard processor subsystem with a programmable fabric infused with AI capabilities. For intelligent edge applications, the FPGA enables real-time compute for time-sensitive applications like autonomous vehicles and industrial Internet of Things (IoT). For smart factory automation technologies like machine vision and robotics, Agilex 3 FPGAs allow for the seamless integration of sensors, drivers, actuators, and machine learning algorithms. To meet the needs of both defense and commercial projects, Agilex 3 FPGAs add several significant security enhancements over the previous generation, including bitstream encryption, authentication, and physical anti-tamper detection. These capabilities ensure reliable and secure performance for critical applications in industrial automation and beyond. Agilex 3 FPGAs leverage Altera's HyperFlex architecture to provide a 1.9 times performance improvement over the previous generation. Extending the HyperFlex architecture to Agilex 3 FPGAs enables high clock frequencies in a power- and cost-optimized FPGA. Additional system performance is achieved through integrated high-speed transceivers operating up to 12.5Gbps and added support for LPDDR4 memory. Software support for Agilex 3 FPGAs will start in Q1 2025 with development kits and production shipments expected to start in mid-2025. Using FPGA software tools to get to market faster Altera also announced the newest features offered in its Quartus Prime Pro software, which helps developers create and compile software faster, improving designer productivity, and accelerating time-to-market. The upcoming Quartus Prime Pro 24.3 release unlocks more devices within the Agilex portfolio and enables improved support for embedded applications. Customers can use this upcoming release to start designing Agilex 5 FPGA D-series, which target an even broader range of use cases compared to Agilex 5 FPGA E-series, which are optimized to deliver efficient compute in edge applications. Altera offers software support for its Agilex 5 FPGA E-series through a no-cost license in the Quartus Prime Software, helping to lower the barriers to entry for Altera's mid-range FPGA family. This software release also includes support for embedded applications that employ either an integrated hard-processor subsystem or Altera's RISC-V solution, the Nios V soft-core processor that can be instantiated in the FPGA fabric. Customers can now access Agilex 5 FPGA design examples that showcase Nios V capabilities such as lockstep, full ECC, and branch prediction. "We've got the software that we we provide to the ecosystem without charge," Rivera said. New OS and RTOS support for the Agilex 5 SoC FPGA-based hard processor subsystem is included in the latest releases of Linux, VxWorks, and Zephyr. Altera and its ecosystem partners announced 11 new Agilex 5 FPGA-based development kits and system-on-modules (SoMs), joining a broad collection of Agilex 5 and Agilex 7 FPGA-based solutions available to help developers get started. FPGA development kits give developers easy and affordable access to Altera hardware, first-hand experience of the capabilities and benefits Agilex FPGAs can deliver, and a faster path to full volume production.

Intel's Altera unit eyes 'audacious' goal to dominate programmable chips

Nine months ago, Intel made Altera -- its unit that produces programmable chips -- a standalone business -- with the intention of spinning it public in 2026. On Monday, Altera executives showed off new chips at their annual developer conference, making the case for why Altera will dominate programmable chips in years to come. Also: Your next laptop may be powered by this Intel AI chip, and the benefits are mind-blowing "Our goal is to be the number one FPGA solutions provider in the world," said Altera CEO Sandra Rivera in a press briefing. "It's a big, audacious, ambitious goal, but it's the right goal for us since we're the only company left in the world that is top to bottom, cloud to edge, FPGAs," said Rivera, referring to "field-programmable gate arrays" -- the programmable chips used across virtually every product in the world that uses chips. Altera, acquired by Intel in 2015 for $15 billion, is one of a triumvirate of programmable chip makers that came to market in the 1980s, the other two being Xilinx, which was acquired last year by Intel's arch-rival, Advanced Micro Devices, and Lattice Semiconductor, which remains independent. The plan, said Rivera, is to take Altera public in 2026, "which is a very fun and important milestone," she said, "but our our journey really is what happens throughout the next number of years on our path to number 1." In reviewing Altera's successes and failures under Intel, Rivera said the company had been even more successful in applications in cloud computing than originally expected. For example, the "infrastructure processing unit (IPU), which can offload tasks such as storage management from the main CPU of a server, was a positive surprise." "That was one we probably didn't see in 2015," said Rivera. "And we actually won every single socket that was available in the market for an FPGA" in IPUs, she added. Also: Making gen AI more efficient with a new kind of chip On the flip side, "Where we fell short is in thinking that we could do more of that co-packaging of the FPGA with the CPU, because we didn't realize at that point how much you would be constraining one device or the other device where the customers really want to use both devices at their full potential. So, that value proposition was never realized, and we moved away from that strategy within the first couple of years." At the developer conference Monday, Rivera and team made product announcements to cement that lead over Xilinx and Lattice. The company unveiled the latest line of programmable chips, the Agilex 3 family of FPGAs. The chip family, which will become available the middle of next year, is designed for cost-sensitive and space-constrained applications in markets such as edge computing and robotics. As Intel details in its press release, "Compared to the previous generation, Agilex 3 FPGAs bring higher levels of integration, enhanced security, and higher performance in a compact package, with densities ranging from 25K-135K logic elements." The chips have greater integration of their parts, said Rivera, including two on-board ARM "Cortex A55" CPUs. That enables the Agilex 3, she explained, "to deliver all of those additional capabilities that you'd need in autonomous vehicles and industrial IoT in this very powerful, small device." The chips also have greater emphasis placed on their internal connective "fabric" to handle real-time computing tasks. That is important, said Rivera, for "time-sensitive networks, obviously, anything that you don't have latency to go round-trip to a centralized data center." The next level up in performance is Intel's Agilex 5, which gets a boost in capabilities with a new "D series" announced at the conference as well. The D series "gives us more capability to address more segments" of the programmable-chip market, said Rivera. Altera has also tripled the number of system boards and reference kits used by partners and resellers of the Agilex chips, noted Rivera. "We're very excited to see our ecosystem growing, investing with us, and growing the availability of these Agilex 5-based design kits, spanning virtually every use case in the market," she said. Also: Police are using AI to write crime reports. What could go wrong? To aid developers, Altera has updated its Quartus Prime Pro software, the main development tool. Among the enhancements are speed-ups in compile time, when a developer's circuit design is downloaded to the programmable chip. "We have leadership compile times," noted Rivera. "Over the course of the products we've introduced in the last several years around this Agilex, platform, we've got almost 30% faster compile times," she said, "which means that the developer can get maybe an extra turn of an algorithm that they're trying to test out. That just makes their development time more efficient, from a three-hour compile time down to a two-hour compile time." The Quartex software also allows designers to implement Altera's CPU-alternative, the "Nios V soft-core processor," which, unlike a hard-wired ARM CPU core, can "be instantiated in the FPGA fabric." Also: The best Windows laptops: Expert tested and reviewed The breadth of capabilities is something Rivera hopes will shine with a newly independent Altera. Under Intel, she said, the focus had narrowed to a small set of customers, which didn't leverage the full potential of the FPGAs. "We weren't as invested in building out that distribution network and all of the capabilities, and getting more reach and scale out of the portfolio through our partners," she said. "And that's because from an Intel strategy perspective, it was less important than the top 25 customers that make up most of Intel's revenue. "And that's something we get to change now because we're committed to the entire breadth of use cases, applications, market segments, and capabilities in the portfolio."

Intel's Altera unit shares new details about upcoming Agilex 3 chips - SiliconANGLE

Intel's Altera unit shares new details about upcoming Agilex 3 chips Altera today shared new details about Agilex 3, an upcoming series of chips optimized to power connected devices and factory automation systems. The Intel Corp. unit first previewed the processor line in February alongside a number of other upcoming products. The hardware reveal coincided with a change to Altera's corporate structure. The Agilex 3 product series comprises specialized processors called field-programmable gate arrays, or FPGAs. In a standard processor, each set of circuits is optimized for a specific task and can't be adapted to other computations. FPGAs, in contrast, contain programmable circuits that developers can configure for a variety of tasks. The chips' customizability enables them to provide better performance than standard processors in certain situations. If a company plans to run an artificial intelligence model on its FPGA, it could optimize the module's programmable circuits for machine learning tasks. Such a configuration would allow the module to run the model faster than an off-the-shelf chip with fewer AI-optimized circuits. Developing a processor from scratch for the use case a company is targeting can unlock even bigger performance improvements. However, designing custom silicon is prohibitively expensive for many hardware projects. FGPAs are used in situations where developing a fully custom chip is impractical, but a company still requires a processor tailored to its specific use case. The FPGAs in the Agilex 3 series are geared towards systems that prioritize power-efficiency over performance. Intel sees the chips finding use in products such as connected devices, industrial robots and autonomous vehicles. The chipmaker revealed today that the Agilex 3 series offers 1.9 times better performance than its previous-generation FPGAs. Each chip in the lineup contains 25,000 to 135,000 logic elements, the parts of an FPGA that can be configured by the user for specific tasks. Agilex 3 processors also contain circuits that can't be reprogrammed, most notably a dual-core central processing unit based on an Arm Holdings plc design. The FPGA series' core processing modules are supported by a number of other components. There are transceivers that can process multimedia files at speeds of up 12.5 gigabits per second. Additionally, Intel has added support for LPDDR4 memory, a type of RAM that is slower than the varieties used in servers but uses less power. Rounding out the enhancements is a set of new cybersecurity features. To prevent hackers from accessing the data being processed by an Agilex 3 chip, Intel has added an encryption mechanism and an authentication feature for verifying user requests. The company says that its upcoming FPGAs can also detect tempering attempts. Alongside the Agilex 3 feature set, Intel today previewed an upcoming release of its Quartus Prime Pro software for customizing FPGAs. The new release supports Agilex 5, a chip line that the company debuted in February alongside the Agilex 3 series. It's based on a 10-nanometer node and offers up to 60% higher performance per watt than some processors from rivals. "Working closely with our ecosystem and distribution partners, Altera remains committed to delivering FPGA-based solutions that empower innovators with leading-edge programmable technologies that are easy to design and deploy," said Altera Chief Executive Officer Sandra Rivera.