Nvidia's DGX Cloud supercomputer is now available in the Oracle Cloud Marketplace.
Oracle customers will now be able to access the Nvidia DGX Cloud AI supercomputing service for training generative AI and other workloads, alongside the Nvidia AI Enterprise software suite.
Each DGX Cloud instance has eight 80-gigabyte Tensor Core GPUs for a total of 640 gigabytes of GPU memory per node. The platform also uses a high-performance low latency network fabric which means that multiple instances can act as one enormous GPU.
The cloud supercomputer offering is based on Nvidia's DGX platform - a dedicated hardware offering that can be purchased and housed in companies' on-premise data centers. Customers can access the DGX Cloud via the Nvidia Base Command Platform, and can instead pay on a monthly basis for their AI workloads.
As the DGX Cloud supercomputer on Oracle is paired with Nvidia's AI Enterprise software, customers can access more than 100 AI frameworks and pre-trained models. The Enterprise software includes the Nvidia NeMo frameworks for building LLMs, Nvidia Rapids for data science and Nvidia TensorRT-LLM and Triton Inferer Server for production AI.
The University of Albany in New York has used the Nvidia DGX Cloud AI on Oracle as the foundation of its AI Plus Initiative which uses AI in fields such as cybersecurity, weather prediction, health data analytics, drug discovery, and next-generation semiconductor design. The University has been using the solution while it builds its own supercomputer.
“We’re accelerating our mission to infuse AI into virtually every academic and research discipline,” said Thenkurussi Kesavadas, vice president for research and economic development at UAlbany.
“We will drive advances in healthcare, security, and economic competitiveness while equipping students for roles in the evolving job market.”
Nvidia announced plans to offer its DGX Cloud cluster in March 2023 noting at the time that Oracle would be the first cloud provider with access. In the future, Microsoft Azure and Google Cloud are also expected to provide the service.
By May 2023, the DGX H100 clusters were being delivered to enterprises wishing to deploy a DGX system at their on-premise data center.
Earlier this month, reports suggested that Nvidia may be looking to aggressively expand into the cloud and was in talks to lease space from a data center operator in order to become a cloud provider of its own.