MS Ai Insider

Revolutionizing AI Infrastructure: Azure’s NVLink 72 Clusters Offer Unmatched Performance and Scalability

Posted by

ailona

–

January 7, 2025

Satya Nadella recently announced the launch of NVLink 72 clusters in Azure, marking a significant advancement in AI infrastructure. With an impressive 900 GB/s bandwidth per GPU, these clusters enable seamless test-time scaling for optimizing large AI models. This development democratizes access to cutting-edge technology, empowering enterprises and researchers to innovate without the need for on-premise supercomputers.2.

Microsoft Unveils NVLink 72 Clusters in Azure

In a recent LinkedIn post, Satya Nadella, CEO of Microsoft, announced the launch of NVLink 72 clusters in Azure. This development promises to revolutionize AI workloads and enhance computational capabilities.

What’s New?

The NVLink 72 clusters are now live in Azure, marking a significant milestone for AI infrastructure. With this launch, Microsoft aims to facilitate the next generation of AI systems.

“Here’s to the next generation of AI built on these systems!” – Satya Nadella

Major Updates

One of the standout features of NVLink 72 is its unprecedented bandwidth. Each GPU can achieve 900 GB/s, ensuring rapid data exchange. This capability accelerates large-scale model training and inference.

Moreover, the clusters support test-time scaling. This allows users to dynamically adjust computational resources, optimizing AI models for real-world tasks.

Azure Integration

Deploying NVLink 72 in Azure democratizes access to cutting-edge infrastructure. This means enterprises and researchers can innovate without needing on-premise supercomputers.

AI Model Breakthroughs

These clusters are particularly well-suited for large language models, generative AI, and reinforcement learning. Their efficiency and bandwidth are game-changers in the AI landscape.

What’s Important to Know

Each rack of NVLink 72 can deliver up to 1.4 exaFLOPS of AI compute power. This translates to performing 1.4×10¹⁸ calculations every second. Such immense power is crucial for complex computations.

“The activation of NVLink 72 clusters in Azure is a pivotal step towards more advanced and scalable AI systems.” – Ganesan Narayanasamy

Additionally, the clusters feature 13.5 TB of high-bandwidth memory per rack. They also provide up to 1800 GB/s of GPU-to-GPU bandwidth, enhancing performance for parallel tasks.

Looking Ahead

With NVLink 72, Microsoft is not just enhancing hardware; it is expanding the boundaries of AI development. The future of AI scaling looks brighter than ever.

As the tech community embraces these advancements, we can expect innovative solutions that will redefine industries and drive further collaboration in AI research.

NVLink 72 clusters deliver up to 1.4 exaFLOPS of AI compute power, equating to 1.4×10¹⁸ calculations per second.

Each rack features 13.5 TB of high-bandwidth memory, enhancing data processing capabilities.

The clusters provide up to 1800 GB/s GPU-to-GPU bandwidth, facilitating rapid communication between GPUs.

NVLink 72 architecture boasts 25 times the energy efficiency compared to its predecessors.

This infrastructure supports advancements in LLM, Generative AI, and multi-modal models, driving innovation across sectors.

From the Stories