Saturday, March 2, 2024

newslınker tv

Top 5 This Week

Related Posts

Amazon Web Services introduces two advanced chips, Graviton4 and Trainium2

In a groundbreaking development, Amazon Web Services Inc. (AWS) has introduced two advanced chips, Graviton4 and Trainium2, during the AWS re:Invent conference keynote, signaling a transformative phase in cloud computing and artificial intelligence (AI) training. These chips are poised to redefine efficiency and performance in their respective domains.

Graviton4: A Leap in Cloud Computing Performance

The Graviton4, an Arm-based processor, emerges as a cornerstone for AWS in offering high-performance and cost-effective solutions for a wide array of cloud compute workloads in Amazon Elastic Compute Cloud (EC2). This processor marks a significant upgrade over its predecessor, the Graviton3, boasting up to 30% enhanced computing power, 50% additional cores, and 75% more memory bandwidth. David Brown, vice president of compute and networking at AWS, highlights Graviton4 as the most potent and energy-efficient chip created by AWS, emphasizing its role in broad workload applications.

AWS’s journey with custom silicon began in 2018 with Graviton1, and each generation has consistently delivered improved performance, reduced costs, and greater efficiency. Today, AWS offers an extensive range of over 150 different Graviton-powered EC2 instance types worldwide, with more than 2 million Graviton processors in circulation.

The Graviton4 processors will power the new memory-optimized Amazon EC2 R8g instance, enabling enhanced performance for demanding applications like high-performance databases, in-memory caches, and big data analytics.

Trainium2: Pioneering AI Training Capabilities

Simultaneously, AWS announced Trainium2, a specialized high-performance chip for training foundational models (FMs) and large language models (LLMs). These models, integral to modern generative AI applications, necessitate processing vast datasets, translating into increased training time and costs. Trainium2 addresses these challenges by offering up to four times the training performance and three times more memory capacity than its first-generation counterpart, along with doubling the energy efficiency.

Trainium2 chips, optimized for deep learning algorithms, will significantly enhance AI and ML workloads, facilitating advancements in natural language processing, computer vision, and recommender models. The Trainium2 will be available in new Amazon EC2 Trn2 instances, which can be scaled up to 10,000 chips in EC2 UltraClusters. This scale promises unprecedented computational power, enabling training of up to 300 billion-parameter LLMs in mere days instead of months.

Impact and Customer Adoption

This leap in chip technology from AWS has not only set new benchmarks in performance and efficiency but also attracted a diverse customer base. Notable clients such as Anthropic, Databricks, Datadog, Epic, Honeycomb, and SAP are already leveraging these new AWS-designed chips.

The introduction of Graviton4 and Trainium2 marks a significant milestone in AWS’s journey of innovation and customer-centric solutions. These chips not only enhance the capabilities of cloud computing and AI training but also reflect AWS’s commitment to pushing the boundaries of technological advancement. As these technologies become more integrated into various industries, they hold the promise of reshaping the landscape of cloud computing and AI, offering more efficient, cost-effective, and powerful solutions for the challenges of tomorrow.

Bilgesu Erdem
Bilgesu Erdem
Bilgesu graduated from Ankara University, Faculty of Communication, Department of Radio, Television and Cinema. After working as a reporter for various television channels and a newspaper, Bilgesu is currently working as a content editor at Newslinker. She loves technology and animals.

Popular Articles