At the Oracle CloudWorld conference which took place today, the Oracle Cloud Infrastructure (OCI) announced the first zettascale OCI Supercluster, accelerated by the NVIDIA Blackwell platform, to help enterprises train and deploy next-generation AI models using more than 100,000 of NVIDIA’s latest-generation GPUs.
Customers can select from a variety of NVIDIA GPUs and install them anywhere with OCI Superclusters, including on-premises, in the public cloud, and on sovereign cloud infrastructure. The Blackwell-based systems, which are expected to be available in the first half of 2019, can scale up to 131,072 Blackwell GPUs with NVIDIA ConnectX-7 NICs for RoCEv2 or NVIDIA Quantum-2 InfiniBand networking, which enables them to provide the cloud with an incredible 2.4 zettaflops of peak AI compute.
NVIDIA GB200 NVL72 liquid-cooled bare-metal instances were also showcased at the show to help power generative AI applications. With the expanded 72-GPU NVIDIA NVLink domain, the instances are capable of large-scale training and inference of trillion-parameter models in real time.
NVIDIA HGX H200, which connects eight NVIDIA H200 Tensor Core GPUs in a single bare-metal instance using NVLink and NVLink Switch, will be available from OCI this year. It can scale to 65,536 H200 GPUs with NVIDIA ConnectX-7 NICs over RoCEv2 cluster networking. Customers who wish to expedite their training workloads and provide real-time inference at scale can order this instance. Additionally, OCI announced that NVIDIA L40S GPU-accelerated instances for midrange AI workloads, NVIDIA Omniverse, and visualization are now generally available.
Oracle’s edge offerings enable scalable AI at the edge, accelerated by NVIDIA GPUs, even in disconnected and remote locations, for single-node to multi-rack solutions. Up to three NVIDIA L4 Tensor Core GPUs, for instance, are now supported in smaller-scale deployments using Oracle’s Roving Edge Device v2. Businesses are utilizing OCI Superclusters with NVIDIA technology to propel AI innovation. Reka, a foundation model startup, is utilizing the clusters to create sophisticated multimodal AI models for enterprise agents.
Dani Yogatama, cofounder and CEO of Reka, stated, “Reka’s multimodal AI models, built with OCI and NVIDIA technology, empower next-generation enterprise agents that can read, see, hear, and speak to make sense of our complex world.” They added, “With NVIDIA GPU-accelerated infrastructure, we can handle very large models and extensive contexts with ease, all while enabling dense and sparse training to scale efficiently at cluster levels.”
