In recent years, the top competition between Intel’s x86 and Arm in the processor field has seen two battles where the competitive situation has fluctuated.One is the globally sensational Apple’s Mac computer chip replacement event, which replaced x86 architecture with self-developed Arm architecture chips; the other is Amazon’s announcement of the Graviton2, based on Arm’s Neoverse N1 server CPU IP, to replace x86 architecture.The two key battles for Arm in the processor field are backed by two giants: Apple and Amazon.Now, Arm is taking advantage of its success and has announced the advancement of Neoverse, adding two new platforms – the Neoverse V1 platform and the second generation N series platform Neoverse N2. Intel has long dominated the server chip fieldIn the server chip field, Intel has long held over 90% market share, and many chip suppliers have repeatedly tried to break this monopoly, but have failed time and again.The most memorable part was Qualcomm’s Arm-based server processor Centriq 2400, which ultimately exited the market.Qualcomm’s return was a significant blow to the Arm camp at the time. However, soon after, Amazon’s Arm-based server chip Graviton, along with the EC2 A1 virtual server based on that chip, emerged, and the market attributed Qualcomm’s failure in the server field to its own issues, leaving Arm still full of hope.In recent years, Arm has also made some adjustments in the server field.Cloud service providers have become Arm’s main targetArm’s success in dominating the mobile market is largely due to its processor IP and the strong ecosystem built with partners, as smartphone brands have invested in developing Arm architecture chips. However, can this successful formula be replicated in the server field?The largest buyers in the server field are cloud platform service providers, including Amazon AWS, Microsoft, Google, Facebook, Tencent, Huawei, Alibaba, Baidu, etc., and these super buyers require an increasing number of servers year by year.Therefore, Arm’s customer base in the server field has begun to shift towards cloud service providers like Amazon and Huawei, and has gained a good reputation.Of course, Arm also has server chip startups like Ampere, representing independent server chip developers.In terms of product strategy, Arm has also launched a new Neoverse brand for infrastructure and server chips to tailor a clearer development path.In the past, Qualcomm and Huawei developed server chips using Armv8 architecture, but the Cortex series is not on the same level as the rival x86 architecture, making it somewhat difficult for Arm to compete directly with Intel using Cortex, so the new Neoverse brand is a suitable positioning and strategy.Chris Bergey, Senior Vice President and General Manager of Arm’s Infrastructure Division, stated that the performance advantages of Arm processor cores have been recognized not only by large internet companies in the US but also by large internet companies in China and have made significant progress in the broader software ecosystem.He further pointed out that Arm places great importance on the software ecosystem in the infrastructure field. Tencent has previously announced that it successfully ported its TARS microservices development framework to Arm CPU architecture; Baidu has also adopted Arm architecture-based intelligent computing in its data centers, which will continuously promote the development of the Chinese infrastructure software ecosystem based on Arm architecture.He emphasized that just as Amazon AWS Cloud has developed a self-research project based on Arm architecture on its cloud servers, it is believed that more and more internet companies in China will be interested in developing cloud server chips based on Arm architecture, and Arm hopes to assist them in delivering similar projects with Neoverse N1 and V1.Arm vs. x86Intel’s x86 architecture has long occupied the server field, inherently possessing a huge advantage, as the software for server systems, power systems, cooling systems, etc., are all designed to work with Intel’s x86 architecture.Furthermore, the advantage of x86 architecture servers is strong performance, but the downside is also high power consumption, which incurs significant electricity costs for data centers each year; in contrast, while the performance of Arm servers may not compare with x86, they can demonstrate advantages in power consumption and cost.For cloud service providers, when server performance is no longer the only consideration, the performance-to-power ratio of Arm servers can stand out.Arm’s entry into the server field aims to change the mindset to encourage more cloud service providers to adopt it and develop more innovative areas rather than competing directly with Intel in existing applications.New Generation Neoverse Series LaunchedWhen discussing the differences between Cortex and Neoverse, Bergey analyzed that Cortex itself is a very broad product series of Arm, while Neoverse is based on Cortex, adding or enhancing features required by the infrastructure and optimizing power consumption and performance for ultra-multi-core devices, including inter-core connectivity and other optimizations.Arm has also launched the second generation N series platform Neoverse N2 and the Neoverse V1 platform to further accelerate infrastructure transformation.Arm pointed out that the launch of the Neoverse solution is Arm’s first step toward infrastructure, driving innovation across various fields from supercomputers to the world’s largest continuously deployed data centers, extending all the way to edge computing. Neoverse V1 Product:The first platform of the V series, compared to N1, its single-thread performance can increase by over 50%.This series is designed toprovide extreme performance, more buffering, caching, allowances, and queuing microarchitecture to accelerate single-thread execution speed.The single-thread performance can increase by over 50%, making it the fastest platform for applications that rely more on CPU performance and bandwidth. Importantly, Neoverse V1 supports the Scalable Vector Extension (SVE), bringing tremendous application potential to high-performance cloud, high-performance computing, and machine learning markets. Additionally, SVE can execute single instruction, multiple data (SIMD) integer, bfloat16, and floating-point instructions based on a software programming model for unknown width vector units, ensuring the portability and longevity of software coding while maintaining efficient execution.
Neoverse N2 Product:Compared to Neoverse N1, single-thread performance improves by 40%, while maintaining the same level of power consumption and area efficiency as Neoverse N1.Neoverse N2 is designed to provide more efficient computing solutions, addressing performance needs across various scenarios from cloud, smart NICs, enterprise networks to low-power edge devices.Arm pointed out that one of the most important goals is to provide the necessary components for partners’ continuous innovation and design flexibility. A key task is Arm’s chip-level interface, which provides opportunities for designing system-level solutions.It is worth mentioning that in addition to promoting Neoverse, Arm is also providing a smooth experience for software developers through the Project Cassini initiative. By establishing standards, platform security, and reference implementations, Project Cassini enables industry partners to deploy usable software on Arm.At the same time, to meet the innovative and flexible design needs of partners, Arm products also support various high-speed connection standards, such as CCIX and CXL, ensuring that its ecosystem can quickly and efficiently launch related technologies.Regarding the launch of the new generation of Neoverse, Arm also pointed out that the launch of Neoverse marks the realization of Arm’s goal over the past decade to build a more efficient infrastructure platform, with the Neoverse computing platform expected to achieve a 30% annual growth in platform performance, continuously optimizing until 2021.
