from the periphery to cloud data centers / ServerNews

from the periphery to cloud data centers / ServerNews


Arm chips are used in a variety of devices, from mobile phones and autonomous vehicles to peripheral systems and the world’s fastest supercomputer, Fugaku in Japan. This wide range of systems is supported by about 13 million developers. Despite the fact that the takeover deal is not fully approved, NVIDIA is actively involved in promoting the ecosystem and offers the developer community an expanding set of proprietary tools.

These include:

  • NVIDIA AI – Industry Standard for Accelerating AI Learning and Inference
  • NVIDIA RAPIDS – a set of software libraries for working with data and analytics on GPUs
  • NVIDIA HPC SDK – compilers, libraries and software tools for high performance computing
  • NVIDIA RTX – Graphics Drivers Providing Ray Tracing and AI Capabilities

For many years, NVIDIA has used Arm architecture in its Jetson SoCs for edge computing and stand-alone machines, and more recently in networking products such as BlueField data processors. BlueField accelerates and secures network and storage for cloud, embedded, and enterprise applications with the support of partners and ISVs.

The transfer of these resources to Arm expands customer choices and opportunities for Arm and NVIDIA partners in four main sectors: high-performance computing (HPC), cloud / data center, edge AI / robotics, and PC.

Arm extends the capabilities of its processor technologies with Scalable Vector Extensions (SVEs) to meet the needs of the world’s largest, most powerful systems. NVIDIA engineers are already working to extend SVE support to the Fujitsu A64FX, which powers the Fugaku supercomputer.

NVIDIA has partnered with Arm and OEM partners in a variety of ways to bring the cloud to gaming. For example, in partnership with Ampere Computing, it expanded its Mt. Jade for cloud gaming. Ampere Altra-based systems include two 80-core Arm SoCs, four NVIDIA T4 GPUs, and an NVIDIA Mellanox BlueField-2 DPU.

Together, they can simultaneously serve 128 gaming users, streaming sessions for Android services in the cloud, popular with the growing number of 5G smartphone users in China. In addition to gaming, cloud service providers are using Arm-powered servers for machine learning, storage, and other GPU-accelerated applications.

The corporate edge is the next big frontier in computing. The NVIDIA EGX edge AI platform will support all major processor architectures, including Arm. It should also be noted that Windows PCs and Chromebooks from a number of vendors already run on Arm processors, and new consumer platforms are coming soon.

The Arm CUDA Toolkit brings NVIDIA GPU-accelerated computing and AI capabilities to the Arm ecosystem. With CUDA, developers have access to TensorRT for deep learning output, DeepStream for video analytics, and more.