The rapid development of artificial intelligence (AI) is transforming the cloud computing landscape, demanding more powerful, efficient, and sustainable solutions. In this context, Arm and Amazon Web Services (AWS) are strengthening their collaboration with the launch of AWS Graviton4 processors, designed to meet the growing demands of modern work environments and AI workloads.
Exceptional Performance with Graviton4
The Graviton4 processors, based on the Arm Neoverse V2 architecture, offer up to 30% higher performance, 50% more cores, and 75% more bandwidthBandwidth is the maximum data transfer capacity… than their predecessor, Graviton3. These improvements have accelerated the adoption of Graviton processors in the ecosystem, driving their use in various cloud applications.
The Neoverse V2 platform integrates innovations such as Scalable Vector Extensions (SVE/SVE2) and support for precision operations like Bfloat16 and Int8 MatMul, significantly optimizing AI workloads, machine learning (ML), and high-performance computing (HPC).
Boosting AI and ML Workloads
With the launch of Arm Kleidi earlier this year, Arm aims to simplify the adoption of AI workloads through specific optimizations for models like Llama 3. These enhancements allow developers to run large language models (LLMs) on Graviton4 with superior tokens per second metrics and reduced times to the first token, without requiring advanced experience in Arm.
Advancements in HPC and EDA Workloads
Graviton4 also represents a leap forward in the HPC realm, providing 16% more memory bandwidth per core and doubling the L2 cache per vCPU compared to Graviton3E. These features are crucial for applications that rely on bandwidth, such as simulation and electronic design automation (EDA). According to Arm’s internal testing, Graviton4 has improved performance by 20% for RTL simulations.
Adoption and Ecosystem Benefits
In recent years, the adoption of Graviton processors has grown exponentially among AWS customers. Companies across various industries are reducing costs, optimizing performance, and enhancing their sustainability metrics by implementing their workloads on these processors. These advantages translate to increased energy efficiency and a lower environmental impact, aligning with global sustainability goals.
Preparations for AWS re:Invent 2024
During the AWS re:Invent 2024 event, over 60 sessions dedicated to AWS Graviton will be presented, highlighting the capabilities of Graviton4 and its impact on the future of cloud computing. The sessions will cover optimizations for AI workloads to benefits for HPC and EDA, showcasing how Graviton processors are revolutionizing the industry.
The Future of Cloud is Arm
With Graviton4 leading the way, Arm reaffirms its commitment to driving the transformation of cloud computing. The collaboration with AWS continues to break down barriers for developers, ensuring they can fully leverage the performance and efficiency advantages offered by the Arm platform. With these innovations, the future of cloud computing looks more promising than ever.