The long wait for a viable Arm server processor finally ended in 2018. The chip design company is hoping 2019 is the year it takes off.
Arm plans to unveil the newest generation of its Neoverse server processor designs on Wednesday, with two versions designed for general-purpose cloud computing and edge computing. The company promised the new designs would deliver a substantial increase in performance over its current generation processor designs, and they were also designed for state-of-the-art chip-making technology that industry giant Intel has had trouble delivering for its own chips.
Arm doesn’t make chips; instead, it designs processor cores that other companies tweak as they see fit to build their own chips. Almost every mobile phone on the planet uses a processor core designed by Arm and built by Apple, Samsung, Qualcomm, or other companies.
They use Arm’s designs in part because they have traditionally been more energy-efficient than Intel’s chips, which still dominate the desktop, laptop, and server markets. But server buyers and sellers have long coveted that energy efficiency in a package that can handle demanding workloads while retaining the flexibility that allows different companies to put their own stamp on the core design.
Amazon Web Services broke through this barrier in a big way late last year, introducing the Arm-based Graviton server processor custom designed by its in-house silicon team at re:invent 2018. We still don’t really have a sense of how many customers are actually interested in running their workloads on the chip, but its debut was a watershed moment for the Arm server community, said Drew Henry, senior vice president for Arm in charge of the company’s server and networking business, in a briefing ahead of the announcement.
“We believe deeply in this growth that’s going to happen in this infrastructure space,” Henry said, emphasizing that for all we talk about cloud and edge computing, a great deal of the world is still running their applications the old-fashioned way.
AWS’s Graviton processor was based on the “Neon” platform within the Neoverse family (code names!), and on Wednesday Arm is unveiling the “N1” platform. Chips based on the N1 designs will be 60 percent faster when it comes to integer performance and 2.5 times more powerful for basic cloud tasks like web serving.
The N1 processor cores were also designed for seven-nanometer processing technology, which is the leading edge of modern chip-making technology. Intel, which otherwise dominates the market for server processors among both cloud vendors and companies managing their own servers, won’t have that capability at its disposal for years after botching the transition to 10nm technology.
The endorsement from AWS and the production problems at Intel have given Arm a huge opening to dent the server processor monopoly that has dominated this market for years, but that potential rests on the backs of Arm licensees like AWS and Ampere to turn designs into real-world products. AMD is also vying for cloud workloads with its Epyc server processor, which enjoys software compatibility with Intel’s designs, creating perhaps the most competitive scenario for server processors in a generation or two.
Arm also took the wraps off its Neoverse E1 processor core Wednesday, which was designed with edge computing and industrial internet-of-things applications in mind. The mobile networking industry is getting ready to build out 5G networks in earnest (which will take a couple of years, misleading commercials aside), and Arm thinks it has the best combination of performance and power efficiency demanded by edge devices and their constrained operating environments out in the real world.
You’ll know that Arm server momentum is real when Microsoft and Google — both companies unafraid of designing their own silicon — unveil cloud computing instances based on Arm chips. It’s pretty likely that everyone in the server world has had a look at Arm’s roadmap over the next three to five years, compared it to Intel’s, and placed some bets on the future of the most fundamental part of cloud computing.