Ampere said its 128-core Altra Max server processor is now available, and the company outlined its roadmap for its next-generation chips.
The chipmaker, which makes server processors based on Arm technology for very large-scale cloud deployments, also said its next-generation products will use cores developed by the company. Ampere chips, manufactured in 5nm, will debut in 2022 and will be followed in 2023.
Ampere’s Altra and Altra Max use Arm’s M1 core. Altra launched its first Ampere processor last year and plans to take on AMD and Intel. Ampere’s argument for cloud deployments is as follows:
- Consistent scaling based on the number of cores added.
- Low energy consumption.
Better performance for workloads such as encryption and multimedia encoding.
Jeff Wittich, product manager at Ampere, said the company can optimize even more with its own cores. “We and our customers benefit from our own hearts. They will be backward compatible, and we will be able to push performance and scale,” he said. “With a 5nm process and telemetry data, there are more ways to manage power and performance and to iterate.”
Additionally, Ampere’s roadmap includes an Ampere Solutions Portal that provides a daily regression update for open-source applications running on the Altra platform.
Ampere’s processors are increasingly adopted by white label server manufacturers who supply hyperscaler cloud providers. Foxconn Industrial Internet, Gigabyte and Inspur are server manufacturers that use Ampere processors for their servers. In addition, Microsoft Azure, Oracle Cloud, Tencent Cloud and Equinix are customers who use Ampere Altra, as well as players such as ByteDance and Cloudflare. “We have 5 of the top 8 cloud providers in the world,” Wittich said.
The launch, roadmap and traction of hyperscaler is part of a larger growth plan, Wittich explained. Once Ampere has a proven track record in hyperscale cloud deployments, it can expand its partnerships with manufacturers of enterprise servers.
“We will focus on on-premises private cloud and not on legacy computing,” said Wittich, ensuring that solutions were being explored.