HPE’s early stab at Arm servers close to a decade ago didn’t pan out, but the company is hoping the second time is a charm.
The company introduced the ProLiant RL300 Gen11 server, which has Ampere’s Arm server processor. The one-socket server is designed for cloud-based applications, with the ability to scale up applications in a power efficient manner. HPE also sells an HPC-focused Arm server, called Apollo 80, based on the Fujitsu A64FX chip.
The Gen11 server has 128 cores and is designed for AI, media, gaming, and applications that typically use the web as a delivery mechanism. It is different from traditional box server environments with x86 chips, which relied on a client-server model.
“This is really about running code in a scalable environment, whatever form that is, typically open source, or modern cloud native software layers, or internally developed platforms and workloads in a containerized environment in a Linux environment,” said Neil MacDonald, executive vice president and general manager for compute at HPE, during a press briefing.
As more applications move to the cloud, a growing number of companies are becoming receptive to Arm designs. AWS has designed its own Arm processor codenamed Graviton, while Ampere’s chips are also offered by Oracle’s cloud service.
MacDonald said HPE is the first top-tier hardware provider to push out an Arm server. The ProLiant RL300 Gen11 server will ship in the third quarter this year.
A lot of Arm server adoption is being driven by the cloud companies that are driving new applications and optimizing workloads, said Jim McGregor, principal analyst at Tirias Research.
The Arm-based ProLiant system is “a Swiss army knife across a range of those software-based scale-out workloads rather than a one-trick pony for a particular specific use case. It’s not a platform for the core of the enterprise with packaged, off-the-shelf classic enterprise software in traditional enterprise IT. That’s not our target, that’s not our goal,” MacDonald said.
This server is different from HPE’s Gen10 server portfolio, which is dominated by x86 chips. HPE will continue selling x86 servers, but the Arm server signals a shift in HPE’s mindset in server products.
“That market needs a new approach from what’s been going on before. Traditional ODM boxes are not going to be the answer,” MacDonald said, adding that the Arm server “gives us an advantage in certain workloads in delivering the performance at lower power usage, with greater cost efficiency as a result for a broad range of those use cases.”
HPE previously tried creating Arm servers close to a decade ago as part of the Moonshot project, with one server using a chip made by AppliedMicro. The Arm server was more of an experiment, and the idea was to bring power-efficient smartphone chips into server environments. The focus at the time was on serving the LAMP (Linux, Apache, MySQL, PHP) stack, but poor software support hindered industry-wide experiments to make Arm servers.
In 2020, HPE introduced its Apollo 80 Fujitsu A64FX-based Arm server, targeted at HPC workloads. Such systems are in use at Los Alamos and Stonybrook University to support scientific simulations.
Now HPE is turning to an Arm server chip from Ampere, which also worked on software compatibility across hundreds of software stacks. Ampere on its website has posted benchmarks for applications on its chips located in Oracle’s cloud service and Equinix’s datacenters.
The Ampere-based server is geared toward service providers and digital-first enterprises focused on developing cloud applications.
“It is a general-purpose microprocessor. It runs Windows, it runs Linux, it does everything. What it is not specifically designed for is legacy,” said Renee James, CEO of Ampere, during a press briefing.
James was previously the president of Intel, and has a lot of experience building x86 servers. She started Ampere Computing as a way to move away from the old x86 designs, and to bring its power-efficient Arm-based chip to hyperscale environments.
“If you have a clean sheet of paper and look forward into where we think the growth in the industry is going to come — from software, from workloads — you need to deliver linear scaling performance elastically. That’s how software that’s developed today consumes performance out of the CPU that is power-efficient so these data centers can continue to grow. You cannot have countries saying no more datacenters because you can’t take any more power,” James said.
The 1U server runs on Ampere Altra and Altra Max processors. The system supports up to 4TB of memory in up to 16 DIMM slots. The system has PCIe Gen4 slots and OCP 3.0 capable slots. It supports NVMe storage, and has 10 SSD slots, and dual M.2 NVMe SSD options for high-performance storage.
The server can also run in OpenBMC environment for Linux environments in hosting services and system providers.