A New Era in Cloud Computing

Cloud computing revolves around performing the maximum amount of work with the minimum amount of time, energy, rack space, and equipment. Arm Neoverse fundamentally changes the performance equation for hyperscale data centers and cloud computing, delivering far more work on far fewer watts than conventional technologies. Designed for managing large numbers of parallel tasks, Neoverse CPUs feature:

 

  • Multiple Cores and Single Threads. Neoverse partners have developed CPUs featuring 128 cores per socket. A single thread design also reduces interference and a better work-per-area ratio.
  • CMN-600 and CMN-700. Our coherent mesh interconnect for linking processor cores and accelerators within an SoCs or chiplets to shared memory resources and I/O.
  • Scalable Vector Extensions. SVE and SVE2 allow developers to deploy and run code ranging from 128 bits to 2048 bits without recompiling, enabling faster time to market and greater choice for customers.
  • Compatibility and Customization. While Arm designs Neoverse CPUs, our partners set the clock speed, I/O, memory subsystem, cache size, and number of cores of Neoverse silicon. The result is greater choice with wide-ranging compatibility.
 

An Expanding Cloud Ecosystem

Cloud service providers and their customers are experiencing the high performance, power efficiency, security, and improved TCO of Neoverse:

Amazon

AWS offers five full instance families for AWS Graviton2 (M6g, C6g, R6g, T6g, X2gd) with nine variations so customers can optimize on cost, performance, memory, or other attributes.

Oracle Cloud

Oracle Cloud Infrastructure has announced its intent to offer flexible VMs and BareMetal services on Ampere Altra processors.

Alibaba

Alibaba and Arm are collaborating to improve the performance of Alibaba’s Dragonwell OpenJKD implementation running on Ampere Altra processors. This collaboration has resulted in a 50 percent improvement in Dragonwell performance on Arm Neoverse.

Tencent

Arm and Tencent are collaborating to expand Arm’s presence in the Tencent cloud with optimization for Java- and C/C++-based applications and the TARS microservices framework.

Microsoft

Microsoft now deploys internal, production-level servers for Microsoft Azure cloud infrastructure based on the Marvell ThunderX2 server processor.

Baidu Cloud

Baidu Cloud leverages the Broadcom Stingray SmartNIC to deliver advanced cloud native applications, network functions virtualization, and distributed security.

Neoverse N2

Raising the Bar for Cloud Performance and Efficiency

Dedicated cores offer a significant uplift in performance over traditional

Neoverse N2 offers an incredible 40 percent IPC performance uplift over Neoverse N1, while retaining industry-leading performance-per-watt efficiency.

Arm Neoverse’s use of dedicated cores provides:

  • Performance consistency – get what you pay for
  • Linear performance scalability – scale-up or scale-out
  • Improved security – no resource sharing between threads
Learn More

Workload-Driven Performance

Every software stack is different. Over the past decade, we’ve collaborated with software leaders such as Red Hat, VMware, SuSE, Canonical, CircleCI and others, to build a comprehensive portfolio of Arm native offerings. The effort shows in the customer benefits:


  • Snap Inc, a popular social media service, was able to reduce fleet size and lower costs for its Kubernetes-based Messaging Core Service.
  • Intuit, a leader in financial services, reduced its costs to run Kafka by 20 percent while maintaining full performance.
  • Honeycomb.io, a cloud performance optimization service, increased performance by up to 30 percent and cut cost by 20 percent.
  • Formula One, the pinnacle of auto racing, sees the potential to lower its costs by up to 40 percent for its CFD simulations.

NGINX Plus

On NGINIX Plus Arm-based M6g instances outperform comparable instances based on traditional architectures by an average of 54%.

Arm Technology and Tools for Cloud Computing

Resources