FPGAs bring power edge AI applications

Patrick Moorhead, President and Founder of Moor Insights & Strategy, said: “Technology trends like 5G connectivity, cloud-based analytics, factory automation and the smart home are driving demand for embedded vision solutions that support machine learning. 

“However, the data latency, cost and privacy issues associated with cloud-based ML analytics have sparked interest among developers in moving more data processing from the cloud to the Edge. But doing so requires OEMs have access to Edge AI/ML inferencing solutions that offer high performance data processing, low power operation, and a small form factor.”

FPGAs are a compelling hardware platform for embedded vision and AI applications, as they perform functions in parallel. This parallel architecture significantly accelerates certain processing workloads, including data inferencing.

Bill Pu, Co-Founder and President, Leopard Imaging, added: “Embedded vision systems are growing in complexity; many systems today use multiple image sensors, displays and cameras. Implementing such systems at the Edge, where device size and power consumption are critical, adds even more design complexity.

“With their very low power consumption, small size, high-performance interfaces, and robust software and IP library, Lattice CrossLink-NX FPGAs let us quickly and easily develop different video signal bridging, aggregation and splitting applications for our industrial and automotive customers using just one device, saving us significant development time and resources.”

The CrossLink-NX family was designed using the new Lattice Nexus platform, which combines a 28 nm FD-SOI manufacturing process with a new, Lattice-designed, FPGA fabric architecture optimised for low power operation in a small form factor.

“Not only does CrossLink-NX lead in power consumption, form factor, reliability and performance in comparison to similar competing FPGAs, it is also supported by a robust library of design software, IP blocks and application reference designs,” said Gordon Hands, Director of Product Marketing, Lattice Semiconductor. “These make it quick and easy for developers to integrate CrossLink-NX FPGAs into new or existing Edge designs.”

Key features of the CrossLink-NX include:

• Low power - built on the Lattice Nexus FPGA platform, CrossLink-NX provides up to a 75% reduction in power consumption compared to competing FPGAs of a similar class.
• High reliability - CrossLink-NX has a Soft Error Rate (SER) up to 100 times lower than similar FPGAs in its class, making it a compelling solution for mission critical applications that must operate safely and reliably. The initial CrossLink-NX device is designed to support ruggedised environments found in outdoor, industrial, and automotive applications.
• Performance - CrossLink-NX delivers enhanced performance enabled by three key elements:
  • Fast I/O support - CrossLink-NX FPGAs are well-suited for embedded vision applications thanks to support for multiple fast I/Os, including MIPI, PCIe and DDR3 memory.
  • Instant on performance – to better support applications where a long system boot time is unacceptable, such as industrial motor control, CrossLink-NX enables ultra-fast I/O configuration in 3ms and total device configuration in less than 15ms.
  • High memory to logic ratio - to efficiently power AI inferencing in Edge devices, CrossLink-NX features 170 bits of memory for every logic cell, the highest memory to logic ratio in its class, providing two times the performance compared to prior generations.
• Small form factor – to support customer system miniaturisation, the first CrossLink-NX device is available in a six by six millimetres form factor which is up to ten times smaller than similar competing FPGAs in its class.
• Software tools and IP – in addition to its new Lattice Radiant 2.0 design software, Lattice offers a robust library of popular IP cores including interfaces like MIPI D-PHY, PCIe, SGMII and OpenLDI and demos for common embedded visions applications such as 4:1 image sensor aggregation.

Originally scheduled for availability in 2020, Lattice is releasing CrossLink-NX ahead of schedule and is already sampling devices with select customers.