Servers: The Right Compute
Showing results for 
Search instead for 
Did you mean: 

Cavium and HPE Fuel Innovation with Disruptive ARM-Based HPC Technology



The ARM HPC ecosystem allows vendors to collaborate and innovate in an open-source community. HPE and Cavium are leveraging ARM-based solutions to optimize HPC workloads and empower users to succeed.

High performance computing (HPC) technologies have empowered organizations in all sectors to operate with unmatched speed and performance for over 50 years. Today, a collaborative ecosystem of vendors is leveraging the Advanced RISC Machine (ARM) architecture to redefine HPC.

ARM-based solutions are enabling change and choice in the HPC market, as well as helping users work faster and more intelligently than ever before. Groundbreaking developments like open compute platforms are equipping users with maCaviumBlog.jpgssive compute capabilities to tackle a variety of workloads, scale for data-heavy applications, and achieve evolving compute goals.

Cavium and Hewlett Packard Enterprise (HPE) have introduced disruptive HPC platform powered by Cavium’s flagship ThunderX2 processor. This processor features a large number of high performance custom ARM cores developed under architecture license from ARM, best-in-class memory bandwidth, memory capacity, and rich IO configurations that provide a major boost in memory bandwidth, drive higher application throughput, and reduce time-to-solution. Together, these leaders are fueling open innovation and pioneering a new generation of HPC capabilities.

The next generation of HPC solutions

ARM processors are designed to address the performance, power, and cost requirements for nearly all application markets. ARM-based SoCs are the most widely used processor architecture on the planet. Originally developed for low-power devices (including most smartphones and tablets), ARM’s architecture license for silicon vendors enables open innovation across a variety of computing verticals, including servers. Processor vendors such as Cavium are delivering full server-class, dual-socket capable high performance SoCs that are optimized for HPC applications. With an extensive commercial and open-source software ecosystem, ARM inspires open innovation to create ideal solutions to satisfy every customer’s needs and requirements.

The 64-bit ARM architecture provides several important benefits for HPC users:

  • Highly integrated SoCs and efficient, high performance cores for greater compute density and power-efficiency
  • Rebalancing of CPU, Memory, and IO
  • End-to-end commercial tools and compiler software suite for building and porting HPC applications
  • Efficient hardware to deliver high performance at lower cost

ARM-based processor offerings allow vendors to optimize their HPC platform for evolving HPC needs. This invites various companies to collaborate and innovate together—from ARM HPC tools, Mellanox InfiniBand, Red Hat Linux Operation System, and SUSE Linux Enterprise, to HPE servers and Cavium processors. The ARM ecosystem fosters digital transformation and differentiation, enabling choice and growth for HPC users. This multi-vendor effort is accelerating ARM software adoption—for embedded processors, all the way to supercomputers—delivering enhanced floating-point performance and higher memory bandwidth to support the world’s most powerful machines.

Introducing ARM-based high performance servers

As a worldwide leader in HPC and innovation, HPE is at the forefront of this competitive, open ecosystem, accelerating a new category of products based on rising customer and industry demands. HPE has gained the support of major HPC vendors, and collectively designed solutions that are enabled and optimized for ARM architectures.

With these goals in mind, HPE introduced the HPE Apollo 70 as its first purpose-built ARM-based server. This compact “4 in 2U” system architecture supports up to four dual-socket ARMv8-A servers (16 DIMMS total) in industry standard 2U server form factor—backed by Mellanox InfiniBand connectors for enhanced IT performance, efficiency, and scalability. Building on the proven HPE Apollo 2000 chassis architecture, the Apollo 70 enables streamlined access to ARM-based HPC:

  • Dense form factor designed for large-scale HPC deployments
  • Integrated design for higher-density while maximizing full system utilization
  • Increased reliability, availability, and manageability designed for HPC
  • Scalable, optimized performance for HPC applications
  • Advanced thermal technology for energy and power efficiency
  • Tier-1 vendor services and support from HPE Pointnext

Comanche, an early-access program run by HPE’s Advanced Technical Group, produced the first iteration of hardware based on Cavium ThunderX2 ARMv-8 processors. Early collaborators received a prototype Comanche system for software development and evaluation. The hardware design was refined twice before being passed to the Apollo team for productization. HPE leveraged software from multiple vendors, including the HPE HPC software portfolio, to develop the final and fastest iteration. This software-optimized hardware provides customers an end-to-end HPC solution that is integrated and validated to handle their most demanding workloads. The design features open standard OCP cards for fabric connectivity, instead of complicating the system by putting a fabric adapter on the motherboard or having to use a high-speed adapter. This allowed HPE to develop the Apollo 70 based on a proven platform while launching with the newest solutions.

With 32 64-bit ARMv8-A custom cores per processor, ThunderX2 is the ideal SoC for a wide range of HPC workloads. The Apollo 70 features two high performance ThunderX2 processors in a dual-socket configuration and allows customers to utilize one, two, or four hardware threads per core, for core counts between 56 and 224 per Apollo, depending on customer workflow. With more cores than an x86 architecture and more memory channels, ThunderX2 delivers high compute FLOPS and 33% higher memory bandwidth to dramatically increase application performance and throughput—and ThunderX2 cores function as one to turbocharge workflow processes and CFD codes. These advancements are vital to HPC performance, giving HPE/Cavium customers the nimblest dense compute platform on the market.

Cavium has been supplying processors for over a decade to 300 customers worldwide, including data centers, storage vendors, telcos, server OEMs, and more. Cavium acquired an ARM license in 2012 and produced the first generation of ThunderX dual-socket capable ARM servers. This was followed by ThunderX2, packed with cores leveraging 3X greater performance, integrated memory channels, IO, and extensive power management.

Cavium is the first vendor to implement a dual-socket ARMv8-A server with a software infrastructure, which reduces time-to-market and drives security to the firmware level. This integration of hardware and software makes ThunderX2 highly differentiated from any other solution on the market, and drawing from the innovative value of the ARM ecosystem.

Inspiring innovation, now and for the future

HPE and Cavium aren’t simply creating products—they are investing in technology for the future. ARM-based HPC solutions are helping customers realize their biggest compute goals. With this new architecture, end users can pursue their own innovations and build out applications that haven’t been approached. And with simple deployment and low TCO, optimization and modernization have never been easier or more affordable.

The ARM ecosystem has opened the floodgates to a new era of compute, and these technologies may be the cornerstone of a true exascale system in the early 2020s.

To learn more about how ARM-based servers are revolutionizing HPC, I invite you to follow me on Twitter at @Bill_Mannel. And for more information on breakthrough innovations like the world-class HPE Apollo 70 and ThunderX2, check out @HPE_HPC and @CaviumInc.

Bill Mannel
VP & GM - HPC & AI Segment Solutions

twitter.gif @Bill_Mannel
linkedin.gif Bill-Mannel

About the Author


As the Vice President and General Manager of HPC and AI Segment Solutions in the Data Center Infrastructure Group, I lead worldwide business and portfolio strategy and execution for the fastest growing market segments in HPE’s Data Center Infrastructure Group which includes the recent SGI acquisition and the HPE Apollo portfolio.