Supercomputing centers around the world looking for a flexible, energy-efficient alternative to diversify their systems are beginning to turn to Arm for their exascale supercomputers.
Arm is the world’s most popular CPU architecture, but the x86 architecture has more than 97 percent of the data center market, thanks to its ecosystem of partners, software and development tools.
NVIDIA has a long tradition of supporting all CPU architectures, including x86 and IBM’s POWER. NVIDIA supports the Arm architecture, too. It’s work we want to greatly expand and accelerate by acquiring the company to create a unique combination for our partners and customers.
“Arm’s energy-efficient, flexible architecture is ideal for all types of workloads, including high performance computing, cloud and edge applications. NVIDIA’s efforts to accelerate Arm applications will fuel innovation, strengthen the ecosystem and provide greater choice to customers and end users,” said Rene Haas, president of Arm’s Intellectual Property Group, in a recent talk at GTC.
Supercomputers Arm for Exascale
Arm’s open architecture gives organizations the flexibility to design CPUs optimized for their workloads. Supercomputing centers around the world are taking advantage of this powerful option.
“As the provider of the Arm-based high performance CPU conceived through the European Processor Initiative to power European exascale supercomputing, SiPearl is committed to advancing the Arm ecosystem,” said Philippe Notton, SiPearl founder and CEO.
“Implementing our CPU in GPU accelerated nodes requires a solid development environment, domain specific libraries, development kits and HPC application tuning. NVIDIA investment and expertise across all of these areas is driving sustained growth of the ecosystem and the market,” he added.
ETRI, a national research institute in South Korea, is developing an Arm-based CPU for their future supercomputer. It will support both double- and mixed-precision math to tackle HPC and AI applications.
And India’s C-DAC is building an Arm-based CPU to power an exascale system for its scientific research and AI initiatives.
NVIDIA Hands Arm New Tools
These are some of many Arm-based HPC initiatives that can take advantage of the NVIDIA HPC software development kit, a comprehensive suite of compilers, libraries and tools that simplify application development and porting to the Arm architecture. The SDK acts as a foundation for an accelerated Arm HPC ecosystem.
The next version of the software — HPC SDK 21.7, coming in July — will deliver more Arm intrinsics, features that compilers can use to tune performance. And it provides customized math functions specifically optimized for Arm CPUs.
In addition, NVIDIA plans to support Scalable Vector Extensions in Arm’s Neoverse platform. SVE first debuted in Fujitsu’s A64FX that powers Fugaku, ranked No. 1 on the TOP500 list of the world’s fastest supercomputers.
Accelerated Arm Kit Coming in July
We’re also making it easier to create, evaluate and benchmark HPC and AI applications on accelerated Arm systems with the NVIDIA Arm HPC Developer Kit. It’s a platform available from NVIDIA and GIGABYTE in the form of software loaded on a server powered by an Ampere Altra Arm-based CPU, NVIDIA A100 Tensor Core GPUs and NVIDIA BlueField-2 DPUs for accelerated networking.
Developers from over 70 leading organizations, including Los Alamos National Laboratory, have applied for early access to the kit that will ship in July.
“We are building some of our key applications along with some AI/ML workflows for the Arm architecture using the NVIDIA HPC SDK,” said Steve Poole, chief architect for next generation systems at Los Alamos. “The NVIDIA Arm Developer Kits will facilitate the transition of our codes to NVIDIA GPUs and Arm CPUs,” he said.
Arming Researchers with Apps
NVIDIA also offers versions for the Arm architecture of popular HPC applications in molecular dynamics (NAMD, Tinker-HP), materials modeling (Quantum Espresso) and other fields.
The apps are available as containers in NVIDIA NGC, a registry of GPU-optimized software. These containers simplify application deployment on Arm-based systems, so researchers can focus on advancing their science.
Extending Arm with More Partners
The power of Arm-based processors combined with NVIDIA GPUs is addressing high performance computing needs beyond supercomputing centers.
For example, NVIDIA GPUs and Ampere Altra CPUs from Ampere Computing are serving markets from data centers to the cloud. In edge computing, NVIDIA is working with Marvell Semiconductor to team its OCTEON Arm-based processors with NVIDIA’s GPUs to speed up AI workloads for network optimization and security.
Our collaborations include systems providers, too. GIGABYTE has been leading the charge in providing accelerated Arm servers.
“We’re excited to be part of the Arm ecosystem with a portfolio of Arm servers powered by NVIDIA’s GPUs and DPUs and Ampere Altra CPUs,” said Etay Lee, CEO of GIGABYTE.
“We’re committed to working with NVIDIA to grow the Arm ecosystem, from the hardware to the software development tools to Arm applications, helping our customers accelerate their HPC workloads across industries,” he added.
More on the Horizon
We’re also expanding the NVIDIA-Certified program of nearly 40 systems from a dozen OEMs to include Arm-powered systems, so organizations have more choice in the kinds of pretested systems they can confidently deploy.
GIGABYTE and Wiwynn plan to offer servers featuring Arm Neoverse-based CPUs with NVIDIA Ampere GPUs, BlueField-2 DPUs or both. The servers are expected to be available next year and will be submitted for NVIDIA certification as they come to market.
We’re just scratching the surface of what’s possible with accelerated, energy-efficient computing fueled by the combination of Arm, NVIDIA and our partners.
To learn more, tune into the NVIDIA ISC 2021 Special Address Monday, June 28, at 9:30 a.m. PT to get an in-depth overview of the latest news from NVIDIA’s Marc Hamilton, followed by a live Q&A panel with NVIDIA HPC experts.
The post NVIDIA, Partners Extending Arm Ecosystem from Exascale to the Edge appeared first on The Official NVIDIA Blog.