FPGA startup wins funds from Xilinx, Samsung

FPGA startup wins funds from Xilinx, Samsung
Business news |
Efinix Inc. (Santa Clara, Calif.), an FPGA startup founded in 2012 by a couple of veteran engineers, has raised $9.5 million in a funding round led by FPGA vendor Xilinx Inc.
By Peter Clarke

Share:

This brings the amount raised by the company to $16 million and the money will be used to bring its first products to market. They use what the company calls ‘Quantum’ programmable technology, a technology the company claims gives it a 4x improvement in power-performance-area (PPA) over traditional programmable architectures.

This is achievable by way of circuit-level architecture implemented on the same SRAM memory cells and baseline process technologies that rival FPGA vendors use.

The technology behind “Quantum” is the use of a universal circuit that can be used as either a routing block or for logic, according to Sammy Cheung, CEO and co-founder of Efinix. Cheung said that traditional FPGAs use dedicated silicon blocks for routing and others for look up tables (LUTs). In contrast Efinix has block that can be configured for routing or logic depending on the circuit being implemented which allows for superior optimization and for a stripping out of complexity that is found in traditional FPGAs.

That reduction in complexity provides Efinix with its 4x advantage Cheung told eeNews Europe.

Sammy Cheung, CEO of Efinix Inc.

The investment round was led by Xilinx and Hong King X Technology Fund (HKX). Others in the round included Samsung Ventures Investment, Hong Kong Inno Capital and Brizan Investments.

“High-volume applications and markets are prime targets for our Quantum-accelerated products,” said Cheung, in a statement. “Combining our Quantum programmable technology and Efinity Integrated Design Environment, we will be launching a number of joint development projects and a new line of silicon product platforms in the coming months thanks to the funding announced today.”

Next: Multiple foundries


The Efinix hardware will initially range from 4K LUTs to 150K LUTs, Cheung said. The company reckons its technology is good for use in mobile products and machine learning applications and it will come to market as both stand-alone ICs and as FPGA fabric licensed for use within SoCs, Cheung said.

Efinix has been working with foundries Semiconductor Manufacturing International Corp. (SMIC) and Taiwan Semiconductor Manufacturing Co. Ltd. and has silicon designs in 40nm and 28nm CMOS process technology. Cheung said the potential area advantage provided by Quantum allows the use of more mature and lower cost process nodes but he added that the company would also be pushing down below 28nm to 10nm “very soon.”

“Efinix’s solution can address a wide variety of applications that are typically not served by today’s FPGAs,” said Salil Raje, senior vice president of software and IP products group at Xilinx. “We are excited to be an investor and look forward to working with them.”

A spokesperson for Samsung Ventures said: “We envision many applications that feature Quantum technology embedded inside ASICs, ASSPs, or FPGAs.”

“Or priority is to do our own branded products but we want to develop a business model with more focus on customer- and application-specific circuits,” Cheung said.

Next: Co-founders


Cheung said that he and his co-founder Tony Ngai, CTO, had formed the company setting out to build a flexible, more efficient FPGA architecture. Conventional FPGA architectures are limited by having to engineer routing blocks for worst-case routing, are not suitable for integration in ASICs and ASSPs because they require special silicon process recipe tweaks.

By moving to a universal block that can be either logic or routing gains an advantage of fine-grained flexibility and optimization, good wide bus performance and flexible pipelining for PPA trade-offs, Cheung said.

“We will get revenue this year from some customers who have prepaid but we are looking to ramp product sales in 2018, Cheung said.

Related links and articles:

www.efinixinc.com

News articles:

High-flying Achronix plans move to ML

FPGA fabric offered for TSMC 16nm FinFET

Western Digital backs processor-in-memory startup

The case for integrating FPGA fabrics with CPU architectures

zGlue fabric reduces size of wearable tech

Linked Articles
eeNews Europe
10s