FPGA Innovation at the Edge: Opening the New Data Path with Lattice Avant-E FPGAs

It’s no secret that Edge computing continues to experience rapid growth as enterprises seek lower latency, and more secure and private processing than centralized cloud datacenters can provide. Billions of Edge compute devices are being deployed in the market, and they’re only getting more advanced with time. Advancements in machine learning and deep neural networks have helped pave the way for this trend, enabling more intelligence to be processed at faster speeds than ever before. However, Edge processing comes with intensive hardware transactions that must be able to continuously perform, so system processors are being pushed to their limits to keep data flows running smoothly.

While SoCs can handle continuous processing, it can come at the expense of higher power consumption and overhead of the OS and other software layers. To address these hardware intensive transactions, system architectures have shifted toward the concept of heterogeneous compute, a way to offload heavy loads from the main processor to other devices sitting right next to the sensor in a system. This type of architecture has become common in Automotive, Robotics, and Industrial systems, in addition to IoTs and PCs. FPGAs offer a parallel compute engine with the ability to optimize hardware blocks to accelerate workloads employed for this purpose.

Lattice is opening a new data path with the launch of the new Lattice Avant™ platform. Lattice Avant is purpose-built to bring the company’s power efficient architecture, small size, and performance leadership to mid-range FPGAs. Offering best-in-class power efficiency, advanced connectivity, and optimized compute, Lattice Avant enables us to address an expanded set of customer applications across the Communications, Computing, Industrial, and Automotive markets.

On December 5, 2022, Lattice introduced its first devices based on this new platform, Lattice Avant-E FPGAs, which are designed to solve key customer challenges at the Edge by combining class-leading power efficiency, size, and performance with an optimized feature set tailored to the needs of Edge applications like data processing and AI.

Avant-E for Edge Processing

Avant-E FPGAs function as an AI engine for Edge devices that can be reprogrammed in the field, allowing system architects to better keep up with the pace of AI innovation than they could with other processing options, like ASICs. The distributed resources on Avant-E FPGAs’ fabric follow a similar architecture to modern AI ASICs, so they can process data in parallel at a lower operating frequency. When compared with similar class competitive devices, Avant-E FPGAs are up to 2.5X more power efficient.

Avant-E FPGAs also feature embedded memory and a single 18 x 18-bit DSP engine that can be configured to up to four 8 x 8-bit DSPs to increase overall processing speeds.

Optimizing the Computer Vision Pipeline

Vision processing is an example of an application requiring acceleration on Edge devices given the intensity of compute required. There are three blocks that intelligence must pass through in the computer vision pipeline: sense, think, and act. Avant-E FPGAs enable the hardware optimized end-to-end implementation of all three key functions.

Sense: Grabbing data from one or multiple sensors so it can be conditioned, or “cleaned,” for intelligent processing.

• Within the pre-processing block, Avant-E FPGAs feature a flexible interface that enables the ease of integration of different types of sensors, allowing for more data to enter the pipeline, faster. The data, or images, can be adjusted and improved within the interface as needed before moving further down the pipeline.

Think: Extracting intelligence from the data using AI.

• The hardware programmability and low power of Avant-E FPGAs help minimize power consumption of intensive computations.

Act: Acting on the processed data.

• Avant-E FPGAs enable hardware accelerated post processing to reduce latency of the overall pipeline, which benefits mission critical applications when the action needs to occur within a given time frame.

Purpose-built for Edge devices with low latency, Avant-E FPGAs are ideal for applications with very specific timing requirements including machine vision in Industrial environments, robotics, robot navigation, and security cameras. For example, automated factory conveyor belts move at high speeds, and the machine vision cameras monitoring products on the conveyor must be able to quickly sense, think, and act on the data fed to it so when a defective product is found, it can be quickly knocked off the line for manual inspection.

A New Level of Performance at the Lowest Power and Smallest Form Factor

Extending Lattice’s low power programmable leadership, Lattice Avant-E FPGAs will enable higher levels of Edge processing performance in existing applications and help drive innovation at the Edge by expanding the types of Computer Vision applications used in the market.

As announced at the livecast launch event, Lattice will introduce more devices in the Avant family to give our customers more ways to unleash their innovation with the Avant platform. Click here to sign up to our newsletter to stay on top of the latest news from Lattice, including when new devices become available.

To learn more about the all-new Lattice Avant platform, visit the Lattice Avant page or visit our website to watch the replay.