by Vinay Ravuri, CEO and Founder of EdgeQ Inc. (edited for clarity by Alan J Weissberger)
EdgeQ emerged from stealth mode in late 2020 as one of the very few semiconductor companies addressing 5G wireless infrastructure with a clean slate design. We wanted to approach 5G as a platform (chip hardware + RAN software) that allows fluidity and scale as customers migrate through new spectrum, new ITU-R standards/ 3GPP specifications, and new endpoints. Our vision is to reconstitute the wireless infrastructure in a manner that is intelligent, agile, and cloud-native.
Removing the Friction Points in 5G:
The success of any silicon design can be premised on three fundamentals: Flexibility, Power, and Cost. 5G is no exception. The industry’s challenge with 5G design can be parsed as follows:
- Flexibility — Very few merchant silicon suppliers can provide a fully production readied L1. With 5G, there is an added complexity where the L1 needs to be adaptive and programmable to the multitude of workloads and use cases.
- Power — At EdgeQ, we are guided by the 50:50 principle. 50% of a base station TCO is related to power. 50% of the power is related to baseband processing. Just changing the power profile of the baseband profoundly impacts the total system.
- Cost – Cost will be a byproduct of the above two. With the anticipated bandwidth and explosion of new end point connections, linear scaling of OpEx and CapEx costs is not sustainable.
Design for Flexibility and Openness: 5G with a Customizable, Deployable PHY:
EdgeQ’s approach is to provide a RISC-V based SoC with a production-grade RAN that is fully customizable via C/C++. This would enable the market with the ability to differentiate 5G at the PHY level. This contrasts with traditional solutions which offer a reference stack for 5G RAN algorithms.
Hardening of the PHY layer is both an intensive and extensive process that typically requires 2-3 years of knowledgeable expertise. The level of integration and development is where most companies struggle or fail. EdgeQ offers an open, production-readied PHY out of the box.
By softwardizing the RAN, we can provide a chip that is dynamically programmable to any of the ORAN splits. New 3GPP standards (like REL16, REL 17, REL 18 upgrade) can be rolled out over-the-air. New features (like URLLC or massive MIMO) can be deployed on the fly. Multiple protocols can be simultaneously instantiated and coherently parallelized. Whether it is 4G, 5G, or AI, each network protocol can be turned on-or-off in the field.
EdgeQ’s “5G Base Station-on-a-Chip” is based on an unique hybrid architecture where a RISC-V compute complex is responsible for baseband processing. Our thesis is to enable the ecosystem access to open-source tools to develop on top of our extended RISC-V instruction sets. For the first time in communications history, the market can capitalize on GNU tool chains (compilers, tracers, debuggers, JTAG ….) for wireless communications.
Design for Low Power: Aggressive 5G Power Management for Wireless Infrastructure:
When EdgeQ instrumented its SoC, all the core functionalities of a base station (baseband processor, host CPU, NPU, FEC Acceleration, NIC, Data Converters, Timing Sync, and Machine Learning) are condensed into a single silicon. Depending on the uploaded PHY firmware, the EdgeQ SoC can be function as a small cell (gNB), or a Radio Unit (RU), or Distributed Unit (DU) – please see illustration below.
This highly integrative approach is particularly meaningful for all-in-one access points, where customers may want to converge all the O-RAN splits onto a single chip, and yet run the entire system over Power over Ethernet (PoE). Equally, with such a low power profile, customers can contemplate new macro base station topologies where densely packed PCIe cards within distributed Units (DU) can be dynamically swapped or “spun up” to scale with real-time performance and functional needs.
Design for Cost: Flexibility + Low Power:
One vector to reducing the overall system TCO is by substantially reducing the power envelop. Through integration, advanced power management techniques, and unique design, all L1 baseband processing functions at EdgeQ run in the low teens, thereby significantly reducing a customer’s TCO.
Secondarily, by creating a fully programmable 5G Base Station-on-a-Chip, features and performance points are now all cloud managed and activated over-the-air via a SaaS subscription model. Operators and service providers can now deploy systems at scale and effect wholesale upgrades without ripping and replacing.
As 5G becomes the “lingua franca” of connectivity, edge demands will impact the fluidity and constructs of 5G infrastructure. Foundational to any wireless infrastructure will be a programmable 5G SoC, capable of adapting to infrastructural demands with a production deployable cellular stack that is customizable. The idea is to present a scalable 5G platform in which customers and developers can leverage open-source tools and C/C++ to configure the chip.
EdgeQ is a leading innovator creating the industry’s first 5G Base-Station-on-a-Chip. Led by seasoned executives from Qualcomm, Intel, and Broadcom who have delivered industry-transforming technologies (4G/5G, WiFi, Wimax, Artificial Intelligence, Cloud Servers) for the last few decades, EdgeQ is inventing a new paradigm in 5G wireless infrastructure.
The company’s vision is to reconstitute wireless infrastructure into a fluid, cloud-based form that would extend robust internet access and communications to remote and dense areas, as well as to the next trillion of interconnected devices.
EdgeQ is backed by world-renown investors, as well as luminary advisors Paul Jacobs, Matt Grob, Ajit Pai, and Rene Hass.
About Vinay Ravuri:
Vinay is CEO/Founder of EdgeQ. He has 25+ years of experience as a distinguished industry executive and general manager for high-tech corporations in compute, communications, and networking. As an executive, he has navigated companies to key strategic inflections in areas of cloud computing, machine learning, and 5G. As Vice President at Qualcomm, Vinay played an integral part in the establishment of its Data Center server and Machine-Learning technology, and drove their advanced mobile software product management. He holds 12 U.S. patents and is a member of the Forbes Technology Council.
Mr. Ravuri obtained a Masters of Engineering (Major: Electrical Engineering Minor: Computer Science) from Georgia Institute of Technology in 1996, and a Bachelors of Engineering from City University of New York in 1994. He also attended Université de Montréal Deep Learning Summer School in 2016.