Flex Logix moves to make FPGA more widely available for edge AI applications
Flex Logix, a semiconductor startup designing reconfigurable AI accelerator chips, is offering customers new options for building edge applications by offering the ability to make chips that are faster and more energy efficient. The company has also decided to offer new options for licensing its technology for use in other chip companies’ products, paving the way for edge AI acceleration to make its way into more products.
Flex Logix recently completed porting and delivered its EFLX embedded 4K eFPGA IP core on TSMC 7 nanometer process technology. The logic and DSP versions allow customers to integrate production-grade application-specific circuits.
With this announcement, companies can develop FPGAs that can be reconfigured even after the tape-out. Flex Logix’s core technology allows customers to change and adapt edge use cases as application requirements shift.
EFLX embedded FPGA is a high-performance solution compatible with most metal stacks. The EFLX 4K logic IP contains 2520 lookup tables in reconfigurable building blocks, 21 Kbit RAM and an improved XFLX interconnect network. Flex Logix’s 2nd generation XFLX programmable interconnect is designed for optimal performance, particularly with arrays.
“Embedded FPGA is a game changer in chip design, particularly to address the rising design complexity at advanced process nodes such as 7nm,” said Andy Jaros, vice president of IP sales and marketing for Flex Logix. “By adding another process node to our portfolio, we can expand application use cases and provide more customers with the flexibility to make changes at any point in the chip’s life span, even in the customers’ systems.”
Flex Logix provides licensing access to InferX AI technology
Flex Logix opened licensing access to its edge inference InferX AI technology. With this access, device manufacturers can take advantage of the excellent performance value for money. The company says they have progressed their InferX technology design to 16nm production in the past four years, outperforming traditional architecture for inferencing.
Flex Logix sees this as an opportunity to increase market adoption of the company’s reconfigurable computing architecture across multiple platforms. Flex Logix’s partners and lead customers can use the company’s AI inference hardware and software IP for SoC integration. Flex Logic anticipates that the AI solution will be available in early 2023.
“As commercial adoption of eFPGA accelerates, with more companies building superior products based on Flex Logix technology, we are pleased to also offer our customers AI IP along with a high-efficiency automated compiler,” said Geoff Tate, co-founder and CEO at Flex Logix. “Flex Logix’s license partners are poised to make reconfigurable computing ubiquitous over the coming decades – through both embedded FPGA and edge inference applications.”
Earlier this year, Flex Logix announced the commercial availability of its EasyVision platform. The purpose of this innovative new platform is to help enterprises bring computer vision products to market faster. The EasyVision platform, featuring the InferX, has ready-to-use AI models pre-trained on large datasets. These models provide users with commonly used object detection capabilities.
Edge Ecosystem Podcast: Kyndryl on integrating the edge ecosystem
Article Topics
chip | edge AI | embedded | Flex Logix | FPGA | programmable
Comments