Introduction
Oracle has achieved a monumental milestone in cloud computing with the introduction of the world’s first zettascale cloud computing cluster.
Powered by NVIDIA’s cutting-edge Blackwell GPUs, this new supercomputer marks a significant leap in AI and high-performance computing.
Overview
- World’s First Zettascale Cluster: Oracle’s new cloud supercomputer features up to 131,072 NVIDIA Blackwell GPUs.
- Unprecedented Performance: The OCI Supercluster reaches an astonishing 2.4 zettaFLOPS, surpassing current supercomputers.
- Flexible Deployment: Customers can choose from various configurations, including NVIDIA H100 and H200 Tensor Core GPUs.
- Real-World Applications: Notable clients such as WideLabs and Zoom are already utilizing OCI’s infrastructure.
- Data Sovereignty: OCI’s cloud infrastructure ensures robust data security and compliance with regional regulations.
A New Milestone in Cloud Computing
Oracle has announced the launch of its zettascale cloud computing cluster, a groundbreaking advancement in the world of AI and supercomputing.
This new cluster is powered by NVIDIA’s Blackwell GPUs and offers unprecedented capabilities. With up to 131,072 NVIDIA Blackwell GPUs, the OCI Supercluster achieves an extraordinary 2.4 zettaFLOPS of peak performance.
This performance level far exceeds existing supercomputers, setting a new benchmark in computational power.
Mahesh Thiagarajan, Executive Vice President of Oracle Cloud Infrastructure, highlighted the significance of this launch:
“We provide one of the most extensive AI infrastructure portfolios, supporting some of the most demanding AI workloads. Our distributed cloud enables unparalleled flexibility in deploying cloud and AI services while maintaining the highest standards of data and AI sovereignty.”
Unprecedented Performance
The OCI Supercluster offers a new level of performance in cloud computing. Here’s a breakdown of its capabilities:
- NVIDIA H100 GPUs: These GPUs allow scaling up to 16,384 units, delivering up to 65 ExaFLOPS of performance and 13 petabytes per second (Pb/s) of network throughput.
- NVIDIA H200 GPUs: Available later this year, these GPUs will scale up to 65,536 units, achieving up to 260 ExaFLOPS of performance and 52 Pb/s of network throughput.
- NVIDIA Blackwell GPUs: Expected in the first half of 2025, these GPUs will use fifth-generation NVLink technology for high-bandwidth communication. They will support up to 72 GPUs with an aggregate bandwidth of 129.6 terabytes per second (TB/s) in a single cluster.
The OCI Supercluster also includes ultra-low latency networking options such as RoCEv2 with ConnectX-7 NICs and NVIDIA Quantum-2 InfiniBand-based networks.
Flexible Deployment Options
Oracle’s zettascale cluster provides a range of configurations to meet diverse customer needs. Organizations can choose from different GPU options, including:
- OCI Compute powered by NVIDIA H100 GPUs: For those needing up to 16,384 GPUs and 65 ExaFLOPS performance.
- OCI Compute powered by NVIDIA H200 GPUs: For higher scalability with up to 65,536 GPUs and 260 ExaFLOPS performance, available later this year.
- NVIDIA Blackwell GPUs: Featuring advanced NVLink technology for high-performance computing and efficient GPU communication.
Real-World Applications
Several leading organizations are already leveraging OCI’s advanced infrastructure:
- WideLabs: A Brazilian AI startup, WideLabs, is utilizing OCI to train Amazonia IA, one of Brazil’s largest language models. Their application, bAIgrapher, helps Alzheimer’s patients by generating biographical content. WideLabs benefits from OCI’s AI infrastructure and NVIDIA H100 GPUs, ensuring that sensitive data remains within Brazilian borders and complies with local AI sovereignty regulations.
- Zoom: The AI-powered collaboration platform Zoom uses OCI to enhance its AI personal assistant, Zoom AI Companion. This tool helps users with drafting emails, summarizing meetings, and brainstorming ideas. OCI’s data sovereignty features enable Zoom to keep customer data localized, particularly in regions like Saudi Arabia, supporting AI sovereignty requirements.
“NVIDIA’s full-stack AI computing platform on Oracle’s distributed cloud will deliver AI compute capabilities at unprecedented scales. This advancement will significantly accelerate global research, development, and deployment efforts.”
Ensuring Data Sovereignty
Oracle’s cloud infrastructure emphasizes data security and regulatory compliance.
The OCI Supercluster’s distributed nature ensures that data can be kept within regional boundaries, meeting local sovereignty and security requirements.
This feature is crucial for businesses and organizations that need to adhere to strict data governance policies.
The Future of AI Supercomputing
Oracle’s introduction of the zettascale cloud computing cluster marks a pivotal moment in the evolution of supercomputing.
The integration of NVIDIA’s Blackwell GPUs into OCI’s infrastructure sets a new standard for performance, scalability, and flexibility in cloud computing.
As the technology advances, the focus will be on maintaining high performance while addressing challenges related to energy consumption and data sovereignty.
The OCI Supercluster represents a significant step forward in AI and high-performance computing, paving the way for future innovations.
Conclusion
Oracle’s introduction of the zetta scale cloud computing cluster is a landmark achievement in the evolution of supercomputing.
With its integration of NVIDIA’s Blackwell GPUs, this new cloud infrastructure sets a new standard for performance, scalability, and flexibility.
As AI and high-performance computing continue to advance, the OCI Supercluster represents a significant step forward, addressing the demands of modern computational tasks and paving the way for future innovations.
