The Markets and Applications for eFPGA
When embedded FPGA (eFPGA) first arrived on the scene, it was long considered the holy grail of chip design. The idea of being able to make a change after RTL was frozen – without having to do another costly and time-consuming spin – was always something sought after but never possible. eFPGA was a game changer in this area because it finally gave designers the flexibility to make changes at any point in the chip’s life span, even in the customers’ systems. This eliminated many expensive chip spins and enabled chip designers to start addressing many customers and applications with the same chips. It also extended the life of chips and systems because designers were now able to update their chips as protocols and standards changed.
eFPGA has come a long way since it first became available about seven years ago. The industry has already seen several generations of products and with each new generation, eFPGA has become more flexible and more usable for new applications, all driven by customer demand. Flash forward to 2021 and the need for reconfigurability in SoCs has never been greater due to the rapidly rising cost of developing SoCs especially at advanced process nodes. In fact, the applications for eFPGA seem to be so endless that this technology is well positioned to become as pervasive as Arm processors have become in the industry.
Today, eFPGA is available from multiple suppliers and has been implemented by customers in nodes from 180nm to 7nm, with 5nm coming up next. There are many publicly announced adopters of eFPGA such as Boeing, Dialog, Harvard, MorningCore, and Sandia National Labs. Many other companies have chosen to keep their applications confidential and there are even more that are currently in evaluation of eFPGA for a wide range of applications.
Next: The markets
The Markets for eFPGA
Below are just a few of the markets and applications ripe to take advantage of eFPGA. Many of these markets were early adopters of the technology and have paved the way for other markets to follow and also take advantage of the financial and time to market advantages this technology offers.
Defense – When one thinks of DARPA, the word innovation typically follows suite. It’s no wonder that DARPA was one of the first vendors to announce support for eFPGA back in 2017. In fact, this program was so successful that DARPA just announced in March 2021 that it had expanded the availability of eFPGA to its research teams. Since research teams typically get funding to tape out a chip once, it made strategic sense to provide them with this technology.
Networking – Smart network interface cards (Smart NICs) often have FPGAs sitting next to the data processing unit (DPU) to support acceleration for encryption/decryption, AI for data analysis, firewall, TCP/IP, HTTP, and compression, etc. With eFPGA, these accelerators can now reside in the eFPGA on the same chip as the Data Processing Unit (DPU), which reduces BOM costs, system power and improves performance. eFPGA has near-hardwired ASIC performance and is also suitable for use in configurable data paths and packet inspection and parsing.
Wireless Communications – Base stations today are major users of FPGAs for flexibility and reconfigurability to allow for customization and real-time updating of protocols and algorithms for applications such as beam forming and networking applications described above. eFPGA allows systems to be smaller, lighter and consume lower power by integrating the FPGA into the application specific integrated circuit/system-on-a-chip (ASIC/SoC).
Aerospace – Boeing was an early adopter of eFPGA back in 2018. Using this technology allows aerospace systems to be smaller, lighter and consume lower power. More importantly, hardware updates can be made to ASICs while orbiting the planet! In addition, eFPGA can be ported to any fab so manufacturing can be in a location preferable to the aerospace company. The technology can even be radiation-hardened if it is designed using a rad-hard standard cell library.
Storage – SSD controllers are putting more intelligence into their systems for better data management to analyze compromised (aka worn) memory location to avoid loss of data, leverage AI for data analysis and update to the latest sorting, encryption and compression algorithms. eFPGA is the ideal programmable solution for these widely different workloads since it can be programmed with any type of hardware accelerator.
Artificial Intelligence – Data centers use FPGA for AI accelerators. eFPGA operates just like FPGA, but within the ASIC. Integrating eFPGA into the main ASIC improves performance by eliminating ASIC to FPGA data latency and system power by removing redundant circuitry that resides on both the ASIC and FPGA such as serdes and PCI PHYs, etc.
MCU/IoT – Cypress (now Infineon) has offered basic reconfigurability with their PSoC MCU product line to provide user defined customization for their application. By incorporating a small amount of eFPGA to Application Specific Standard parts (ASSPs) and application processors, chip suppliers can provide their customers the ability to customize their chips while producing a standard product. eFPGA is an ideal security solution for IoT applications because system companies can keep their proprietary circuit design in-house and program the MCU with eFPGA post-manufacturing.
The use of eFPGA is spreading across many market segments as an alternative programmable solution to embedded processors, DSPs and GPUs. Its flexibility, ability to support more compute parallelism and end-user customization post-silicon will become a mainstream requirement for systems companies.
Next: The future
eFPGA is a very powerful and flexible technology, applicable to a wide range of markets and applications. It will continue to evolve as customers learn how to use eFPGA and continually ask suppliers to support new features and capabilities to improve its value proposition. Real customer needs will be the driving force behind the future evolution of eFPGA and it will be exciting to see how this technology continues to be evolved over time.
Andy Jaros, vice president of IP sales, marketing and solutions architecture, for Flex Logix, is responsible for the EFLX and nnMAX IP. Jaros started in the electronics industry as a sales manager with Motorola Semiconductor, moved to ARM as director of strategic accounts and then to ARC as vice president of sales North America. Prior to joining Flex Logix Jaros was director of product solution sales for all ARC sales activities for Synopsys in North America.
Related links and articles:
Embedding an eFPGA: the starting point
The case for integrating FPGA fabrics with CPU architectures
Flex FPGA ported to GloFo’s FDSOI process
Flex Logix licences FPGA fabric to China’s MorningCore
Sifive signs Flex Logix for low-cost FPGA fabric
Quicklogic turns to licensing FPGA fabric