UPDATED 15:33 EDT / DECEMBER 03 2019

CLOUD

AWS impresses with its latest line of custom-made cloud processors

For a company that started out selling books, Amazon.com Inc. is demonstrating that it knows a thing or two about microprocessors as well.

The e-commerce giant’s cloud computing arm today rolled out the next generation of the microprocessors that will power some of its mainstream Elastic Compute Cloud or EC2 instances. Announced at Amazon Web Services Inc.’s annual re:Invent conference this morning, the Graviton2 processor was designed by AWS based upon a reference architecture from Arm Holdings Plc.

Built using a seven-nanometer manufacturing process, it delivers up to seven times the performance of the current Graviton-powered A1 instances with twice the floating-point performance, more memory channels and up to five times faster memory access. It’s more secure as well, with memory protected by 256-bit encryption.

AWS said the chips will power its M6g, C6g and R6g instances, which are versions of EC2 tuned to handle different kinds of workloads. Customers will see up to 40% better price/performance than they do currently. The new chip is already powering M6g instances, which are intended to support a broad variety of workloads. Support for the more specialized compute- and memory-optimized instances is planned for next year. Customers will be able to choose up to 64 virtual CPUs, 512 gigabytes of memory and 25 gigabits-per-second of networking.

The announcement was expected, but the specs blow away the rumored 20% performance improvements. “AWS is dead-serious with this new chip, and I don’t say that lightly,” said Patrick Moorhead, president and principal analyst at Moor Insights & Strategy.

The more muscular chips could force other cloud providers to go deeper with their own strategies based on Arm and processors from Advanced Micro Devices Inc., fueling a performance war in which Moorhead said customers will be the winners. “Whatever the outcome, we now have even more competition in the data center processor market,” he said.

Secret weapon

The new processors are the latest innovation from Annapurna Labs Ltd., a secretive Israel-based chip foundry that Amazon acquired four years ago. Annapurna designed the original Graviton, which debuted last year. In addition to the Graviton2, it was also the brains behind Inferentia, a machine learning-optimized processor that AWS announced last year and that it announced today is now in use in Inf1 instances targeted at machine learning.

In contrast to chips aimed at the training stage of machine learning, Inferentia is intended to speed the process of makes inferences, or predictions, which the company claims are the most computer-intensive part of machine learning models. CEO Andy Jassy said Inferentia delivers the “fastest inference in the cloud” with three times the throughput of current G4 instances that use the popular Nvidia Corp. graphic processing units at 40% lower cost per inference.

Even with the latest performance boost, Graviton2 isn’t expected to challenge the most powerful processors from Intel Corp. or AMD, but that isn’t necessarily the point. Arm processors are up to 45% cheaper to produce and are famous for their stingy power consumption. That means AWS can pack more computing power into the same space than competitors that use off-the-shelf processors.

The result should be “beneficial for both AWS and its cloud customers,” said Charles King, principal analyst at Pund-IT Inc. “Better price/performance means improved customer experience and profit margins.”

Shot in the Arm

But King stopped short of calling the announcement a game-changer, noting that Arm-based chip development “has seriously lagged expectations for years. I don’t expect Graviton2 to make much of a dent in Intel’s position, at least in the short term,” he said.

For one thing, Graviton is currently limited to supporting only Linux and Unix, King said, and it’s unclear whether customers who use instances based upon the architecture will be locked in to doing business only with AWS. “Silicon-based vendor lock-in was a common theme during the heydays of [reduced-instruction-set-computing ]-based systems,” he said. “I don’t think many enterprises will be enthusiastic about returning to that business model.”

A more important message to the market may be that Amazon is serious about becoming a major force in hardware design, Moorhead said. Graviton2 is debuting just 10 months after the reference architecture was released by Arm, a time frame the analyst said is “literally unheard of. Even assuming AWS got the intellectual property early, which I’m sure they did, this is about as agile a silicon development lifecycle as one can imagine.”

Photo: Robert Hof/SiliconANGLE

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU