Oracle Cloud Infrastructure with Nvidia A100 tensor core GPUs on bare metal instances is said to be able to run complex AI models and deep learning systems between two-and-a-half to six times faster than instances featuring previous generations of GPUs.
Oracle Cloud also allows customers to run complex HPC applications using GPUDirect over Nvidia Mellanox RDMA networking for reduced latency.
The new GPU4.8 bare metal instance features eight Nvidia A100 Tensor Core GPUs with 40 GB of memory each, all interconnected via Nvidia NVLink. The onboard CPU on board has 64 physical 2.9GHz AMD Rome cores, 2,048GB of RAM and 24TB of NVMe storage.
Virtual machines with one, two, or four GPUs per virtual machine will be available in the coming months.
"Since its inception, Oracle Cloud Infrastructure has standardized on Nvidia's GPUs, beginning with the Pascal generation, moving to Volta and now with Ampere. Our customers demand the best of on-premises with all the benefits of the cloud, which is what we're delivering with our latest GPU instance running on Nvidia's latest A100 GPU," said Oracle Cloud Infrastructure vice president Karan Batta.
"We have the largest, most performant, and most cost effective A100 offering in the cloud because we offer double the memory and more local storage than competitors. This is the GPU instance customers have been waiting for to move to the cloud and deliver important breakthroughs."
Nvidia general manager and vice president of accelerated computing Ian Buck said "Accelerated computing is essential to driving research breakthroughs for enterprises across all industries," said .
"By bringing the Nvidia A100 Tensor Core GPUs into its cloud service, and offering the ability to scale to more than 500 GPUs interconnected with Mellanox networking, Oracle is providing the computing performance needed to accelerate the most critical work being done today in AI and high performance computing."
Oracle also announced images, stacks and services that make it easier to run certain kinds of workload on Oracle Cloud. They include:
AI Engineered MLOps Solution Stack to automate the instantiation of a cluster of bare metal Nvidia A100 shapes for distributed training and inference.
Media Ops Packaged Solution Stacks to automate the operation of self-managed media operations pipelines.
Cloud Native MLOps Orchestration Package to automate the movement of data and the creation of compute instances.
Julia AI HPC Image is an engineered HPC Image with Julia, Jupyter Notebook and IJulia, tested and optimised for Nvidia A100 GPUs. .
Pre-configured Data Science and AI Image includes Nvidia's Deep Neural Network libraries, common ML/deep learning frameworks, Jupyter Notebooks and common Python/R integrated development environments.