Enterprises are looking for increasingly powerful computing to support AI workloads and accelerate data processing. Greater efficiency translates into better returns on investments in training and fine-tuning AI, and a better user experience with AI inference.
Today at the Oracle CloudWorld conference, Oracle Cloud Infrastructure (OCI) announced the first Zetascale OCI supercluster accelerated by the NVIDIA Blackwell platform, enabling enterprises to train and deploy next-generation AI models using more than 100,000 of NVIDIA’s latest generation GPUs.
OCI Superclusters enable customers to choose from a wide range of NVIDIA GPUs and deploy them anywhere – on-premise, on public cloud or sovereign cloud. Available early next year, Blackwell-based systems can scale up to 131,072 Blackwell GPUs using NVIDIA ConnectX-7 NICs for RoCEv2 or NVIDIA Quantum-2 InfiniBand networking, delivering an astounding 2.4 zetaflops of peak AI computing to the cloud. (Read the press release to learn more about OCI Superclusters.)
At the show, Oracle also previewed the NVIDIA GB200 NVL72 liquid-cooled bare-metal instance to power generative AI applications. The instance enables large-scale training using Quantum-2 InfiniBand and real-time inference of trillion-parameter models within an expanded 72-GPU NVIDIA NVLink domain that can act as a single large GPU.
This year, OCI is offering the NVIDIA HGX H200, which connects eight NVIDIA H200 Tensor Core GPUs to a single bare-metal instance via NVLink and an NVLink switch, scaling to 65,536 H200 GPUs using NVIDIA ConnectX-7 NICs over a RoCEv2 cluster network. This instance is available to order for customers looking to deliver real-time inference at scale and accelerate training workloads. (Read our blog on OCI Superclusters with NVIDIA B200, GB200, and H200 GPUs.)
OCI also announced the general availability of NVIDIA L40S GPU-accelerated instances for mid-range AI workloads, NVIDIA Omniverse, and visualization (read our blog about OCI Superclusters with NVIDIA L40S GPUs).
From single-node to multi-rack solutions, Oracle’s Edge products deliver scalable edge AI accelerated by NVIDIA GPUs, even in remote, disconnected locations. For example, small deployments with Oracle’s Roving Edge Device v2 now support up to three NVIDIA L4 Tensor Core GPUs.
Enterprises are using NVIDIA-powered OCI superclusters to drive AI innovation, such as foundational model startup Reka, which is using the cluster to develop advanced multimodal AI models and develop enterprise agents.
“Reka’s multimodal AI models built with OCI and NVIDIA technology enable the next generation of enterprise agents that can read, see, hear and speak to understand a complex world,” said Dani Yogatama, co-founder and CEO of Reka. “The NVIDIA GPU-accelerated infrastructure enables us to easily handle extremely large models and broad contexts, while efficiently scaling dense and sparse training at the cluster level.”
Accelerating Generative AI Oracle Database Workloads
Oracle Autonomous Database adds NVIDIA GPU support for Oracle Machine Learning Notebooks, enabling customers to accelerate data processing workloads on Oracle Autonomous Database.
At Oracle CloudWorld, NVIDIA and Oracle are partnering to demonstrate three capabilities that show how NVIDIA’s accelerated computing platform can be used now or in the future to accelerate key components of a generative AI search augmentation generation pipeline.
In the first session, we will show how NVIDIA GPUs can be used to accelerate large volumes of vector embeddings directly within Oracle Autonomous Database Serverless, efficiently bringing enterprise data closer to AI. These vectors can then be searched using AI Vector Search in Oracle Database 23ai.
In the second demonstration, we present a proof-of-concept prototype that uses NVIDIA GPUs, NVIDIA RAPIDS cuVS, and an Oracle-developed offloading framework to accelerate vector graph index generation, significantly reducing the time required to build an index for efficient vector searching.
The third demonstration will show how NVIDIA NIM, a set of easy-to-use inference microservices, can improve generative AI performance for text generation and translation use cases across a range of model sizes and concurrency levels.
These new Oracle Database features and demonstrations combine to highlight how enterprises can leverage NVIDIA GPUs to apply generative AI to structured and unstructured data stored or managed in Oracle Database.
Sovereign AI around the world
NVIDIA and Oracle are working together to deliver unique AI infrastructure at global scale to address the data residency needs of governments and enterprises.
Brazil-based startup Wide Labs used NVIDIA H100 Tensor Core GPUs and the NVIDIA NeMo framework in OCI’s Brazilian data center to train and deploy one of the first large-scale language models for Brazilian Portuguese, Amazonia IA, helping to ensure data sovereignty.
“Developing its own LLM gives Amazônia a unique market position, allowing it to offer its customers services processing data within Brazil,” said Nelson Leoni, CEO of Wide Labs. “Using the NVIDIA NeMo framework, we successfully trained the Amazônia IA.”
In Japan, Nomura Research Institute, a leading global provider of consulting services and system solutions, is using OCI’s Alloy infrastructure and NVIDIA GPUs to power its financial AI platform with LLM, which operates in accordance with financial regulations and data sovereignty requirements.
Communications and collaboration company Zoom will use NVIDIA GPUs in OCI’s Saudi Arabian data centers to help comply with local data requirements.
And geospatial modeling company RSS-Hydro is demonstrating how its flood mapping platform, built on the NVIDIA Omniverse platform and powered by L40S GPUs on OCI, uses a digital twin to simulate the impacts of flooding in Japan’s Kumamoto region to mitigate the effects of climate change.
These customers are among the many countries and organizations building and deploying domestic AI applications powered by NVIDIA and OCI to drive economic resilience through their own AI infrastructure.
Enterprise-Ready AI with NVIDIA and Oracle
Enterprises can accelerate task automation on OCI by using OCI’s scalable cloud solutions to deploy NVIDIA software such as NIM microservices and NVIDIA cuOpt. These solutions enable enterprises to rapidly adopt generative AI and build agent workflows for complex tasks such as code generation and route optimization.
NVIDIA cuOpt, NIM, RAPIDS and more are part of the NVIDIA AI Enterprise software platform, available on the Oracle Cloud Marketplace.
Learn more at Oracle CloudWorld
Join NVIDIA at Oracle CloudWorld 2024 to learn how our collaboration is bringing AI and accelerated data processing to organizations around the world.
Register for the event to watch sessions, see demos, and join us for the Oracle and NVIDIA solutions keynote, “Maximizing AI Performance with the NVIDIA Accelerated Computing Platform” (SOL3866), on Wednesday, September 11 in Las Vegas.
