ARM server chips get a boost with Ampere’s Altra – The New Stack

Ampere Computing offers ARM-based chips the kind of server breakthrough that has eluded other chip companies for the past decade.

The company’s chips over the past month have been adopted by Google Cloud and HPE, whose support provides a big boost to the entire ARM server ecosystem.

Ampere was founded by former Intel chairwoman Renee James, who left Intel in 2016 after a leadership shakeup. She previously ran Intel’s data center business and created Ampere after realizing that Intel’s x86 chips wouldn’t be enough for cloud-centric workloads like web applications.

“Where we think industry growth will come from, software, workloads… you have to deliver linear scaling performance elastically. This is how software developed today consumes power-efficient processor performance,” James said during a press briefing last month.

Ampère’s fastest Altra chip has 128 cores and is based on the company’s custom ARM architecture. That’s twice the number of cores of AMD’s high-end x86 server chip and more than three times that of Intel’s x86 server chip, which has 40 cores.

Focus on servers

ARM and x86 chips for servers are fundamentally different with contrasting origin stories. ARM architecture emphasizes power efficiency with roots in smartphones, while x86 server chips are designed for performance and are descendants of power-hungry chips originally designed by AMD and Intel for applications such as databases and ERPs.

“[Altra] is a general purpose microprocessor. It runs Windows, it runs Linux, it does it all. What it’s not specifically designed for are legacy applications,” James said, later adding “I have a lot of experience building x86 servers.”

Vendors adopting Ampere chips have pointed out that Altra cores offer more performance while consuming less power than x86 chips, representing more savings.

Google Cloud announced last week that it would use Ampère’s processors in Tau T2A virtual machine instances, which are designed for video, databases and AI. The T2A with Ampère’s chips provides 31% more power than its x86 counterparts while saving up to 65% on price and performance, according to Google.

Last month, HPE launched its next-generation Proliant Gen11 server line with Ampere’s Altra chip. For decades, HPE has been introducing new generations of single- and dual-socket servers based on x86 chips, but decided to go the ARM route with the new Proliant RL300 servers, which also signals a big shift in how HPE is considering cloud servers. On the high-end computing side, HPE already uses an ARM-based Fujitsu A64FX processor in its Apollo 80, which is aimed at high-performance systems with eight unique sockets in a 2U footprint.

“It’s really about running code in a scalable environment, whatever form it takes – open source, modern cloud native software, layers or in-house developed platforms and workloads in a containerized environment under… Linux,” Neil MacDonald, executive vice president and general manager of HPE’s IT business, said during a press briefing.

Past attempts

A decade ago, HPE’s experiment with ARM chips, an effort called Moonshot, failed. At the time, many startups were trying to bring ARM chips in smartphones to servers, focusing on the LAMP stack (Linux, Apache, MySQL, Perl/PHP/Python).

A lack of interest from server makers in ARM designs led to the closure of a few companies such as Calxeda, while other efforts by companies like Cavium were abandoned as the software ecosystem was immature. Most ARM software development at the time was focused on mobile devices, while server software was focused on x86.

But the scene is very different now, MacDonald said.

The new World

“What has changed – if you look at the ecosystem…there is a strong chain of mature developer tools that has been exacerbated and accelerated by other industry initiatives that have matured. Open source and broader software packages of the past five years are more at a point where there are other choices from a platform development standpoint than there were in the past,” MacDonald said.

Ampere has been particularly focused on software development to boost adoption of its chips so as not to repeat the mistakes of companies like Calxeda, which bit the dust despite its groundbreaking work in hardware.

“There have been many pioneers who have tried to create alternatives in the industry. And it was very difficult to build anything better than what was designed for building client-server computing, which was x86. It was a perfect tool it was built for. It’s not the perfect tool for cloud natives,” James said.

Ampere has released benchmarks for cloud, database, and other popular native applications. It also tests on systems with its chips on services like Google Cloud to ensure reliability and performance.

“We have nightly regressions on hundreds of different software stacks that people use on our products,” James said.

let the games begin

But Ampere’s initial success also exposes it to competition from big brands like Amazon, which has built its own ARM-based chip called Graviton, which it uses on AWS.

ARM architecture will have a 22% share of data center servers by 2025, according to a study by Trendforce. The expansion will be led by AWS’ Graviton, which is now in its third generation.

AWS deployment of ARM-based processors in 2021 reached 15% of global server deployment and will exceed 20% in 2022,” TrendForce said in a study published in March.

HPE’s MacDonald said ARM is an addition to its portfolio and will continue to offer a wide range of chips, meaning the company will continue to offer x86 chips. HPE used AMD’s x86 chips in the supercomputer called Frontier, which was the world’s fastest supercomputer in the latest Top500 list released last month.

x86-based chips remain a mainstay for cloud providers. But growing competition from ARM is one of the reasons Intel and AMD are leaning forward to customize chips for large customers, said Jim McGregor, principal analyst at Tirias Research.

“Cloud providers will manufacture chips themselves, like with the Graviton processor, or Google with its TPU accelerator. Or they will work with a partner. And if you want to be a partner, you have to be flexible. I think Intel is adapting to the new world. The new world is that you have to be flexible with your client and the workloads they’re running,” McGregor said.

Comments are closed.