Arm is and always has been successful in the data center infrastructure market. And this, not only in terms of the cloud, but in all that concerns data centers in companies. It is on this momentum in particular that Arm recently announced Neoverse V2, its updated IP wallet.
A quick introduction to Neoverse
Before starting the announcement made by ARM, it is nevertheless useful to recall Neoverse. The data center consists of three categories of workloads. The first is for performance-intensive workloads. Here, every microsecond of latency counts and every bit of compute resource is useful. Therefore, High Performance Computing (HPC), Artificial Intelligence/Machine Learning (AI/ML) and Deep Analytics are concrete examples. By the way, in these cases, the organization will consume every watt of energy to achieve the best performance.
The second category is scalable workloads. These are distributed over many nodes. As a result, the need for computing resources per node is much reduced. This is for example the case of native cloud applications based on containers. They are actually made up of hundreds of services running from top to bottom.
The third category is networking functions that move data north, south, east, and west. The latter also secure this data. Each category has unique characteristics that translate into silicon (CPU) designs. Therefore, sometimes use cases claim powerful cores tied to lots of memory and cache. In others, less efficient cores and a lot of throughput are required to move data packets.
ARM officially announces Neoverse V2: its updated IP wallet
Neoverse was born in 2018. Since then, ARM has gained a foothold in most CSPs or major cloud service providers around the world. However, CSPs are not content with ARM-based instances. As a result, they decide to extend the services, but on condition that customers demand more.
However, ARM’s Neoverse V2 model is undeniably powerful. In particular, it has the ability to provide an architecture and an IP address. Therefore, this force is exploitable and visible through the different chip implementation methods. The first is through chip vendors like NVIDIA and Ampere. The latter supply the Altra and Altra Max processors used by CSPs and OEMs. Regarding the second, it involves the development of custom SoCs from CSP such as AWS (Graviton) and Alibaba Cloud (Yitian) for specific needs. Besides, server vendors like Fujitsu are also designing another one. The goal here is to generate absolute HPC performance with the A64FX chip. And finally, Neoverse V2 is used for accelerators and other network functions. This last feature affects Intel (Mt Evans), AMD (Pensando), Marvell, NVIDIA and many others.
We would like to say thanks to the writer of this article for this remarkable material
Arm continues its data center and launches Neoverse V2 – LeBigData.fr
Explore our social media accounts and other pages related to themhttps://www.ai-magazine.com/related-pages/