Go to our on-demand library to view VB Transform 2023 sessions. Register here
Nvidia today announced the broad accessibility of its cloud-based AI supercomputing service, DGX Cloud. This service will grant users access to thousands of virtual Nvidia GPUs on Oracle Cloud Infrastructure (OCI), along with infrastructure in the US and UK
DGX Cloud was announced during Nvidia’s GTC conference in March. He promised to provide enterprises with the infrastructure and software needed to train advanced models in generative AI and other fields that use AI.
Nvidia said the purpose-built infrastructure is designed to meet gen AI demands for massive AI supercomputing for training large and complex models such as language models.
>>Follow VentureBeat’s continued coverage of generative AI<
“Just as many enterprises have deployed on-premises DGX SuperPODs, DGX Cloud leverages the best computing architecture, with large clusters of dedicated DGX Cloud instances interconnected over Nvidia’s ultra-high-bandwidth, low-latency network fabric,” Tony Paikeday, senior director, DGX Platforms at Nvidia, told VentureBeat.
Paikeday said DGX Cloud simplifies the management of complex infrastructures, providing an easy-to-use “serverless AI” experience. This allows developers to focus on running experiments, building prototypes, and reaching viable models faster without the burden of infrastructure hassles.
“Organizations that needed to build generative AI models before the advent of DGX Cloud would have had only on-premises data center infrastructure as a viable option to address these large-scale workloads,” Paikeday told VentureBeat. “With DGX Cloud, any organization can now remotely access their AI supercomputer to train large and complex LLMs and other generative AI models from the comfort of their browser, without having to manage a supercomputing data center.”
>>Don’t miss our special issue: The Future of the Data Center: Handling ever-increasing demands.<
Nvidia says the offering allows AI developers to deploy heavy workloads across multiple compute nodes in parallel, leading to training speeds two to three times faster than traditional cloud computing.
The company also says DGX Cloud allows companies to establish their own “AI center of excellence,” supporting large teams of developers working on numerous AI projects simultaneously. These projects can benefit from a pool of supercomputing capacity that automatically caters to AI workloads as needed.
Streamlining enterprise generative AI workloads using DGX Cloud
Second McKinseyGenerative AI could contribute more than $4 trillion annually to the global economy by transforming proprietary business knowledge into next-generation artificial intelligence applications.
The exponential growth of generative AI has forced leading companies across various industries to adopt AI as a business imperative, fueling the demand for accelerated information infrastructure. Nvidia said it has optimized DGX Cloud’s architecture to meet these growing computational needs.
Nvidia’s Paikeday said developers often face challenges in preparing data, building initial prototypes, and using GPU infrastructure efficiently. DGX Cloud, powered by Nvidia Base Command Platform and Nvidia AI Enterprise, aims to solve these problems.
“Through the Nvidia Base Command Platform and Nvidia AI Enterprise, DGX Cloud enables developers to get production-ready models sooner and with less effort, thanks to accelerated data science libraries, optimized AI frameworks, a suite of pre-training AI models, and workflow management software to speed model creation,” Paikeday told VentureBeat.
Biotechnology company Amgen use DGX Cloud to accelerate drug discovery. Nvidia said the company uses DGX Cloud in conjunction with Nvidia BioNeMo large language model (LLM) software and Nvidia AI Enterprise software, including Nvidia RAPIDS data science acceleration libraries.
“With Nvidia DGX Cloud and Nvidia BioNeMo, our researchers can focus on deeper biology instead of having to deal with AI infrastructure and configure ML engineering,” said Peter Grandsard, executive research director, Biological Therapeutic Discovery, Center for Research Acceleration by Digital Innovation at Amgen, in a written statement.
A healthy case study
Amgen says it can now rapidly analyze trillions of antibody sequences via DGX Cloud, enabling the rapid development of synthetic proteins. The company reported that DGX Cloud’s compute and multi-node capabilities enabled it to train protein LLMs three times faster with BioNeMo and up to 100 times faster post-training analysis with Nvidia RAPIDS compared to alternative platforms.
Nvidia will offer DGX Cloud instances on a monthly basis. Each instance will feature eight powerful 80GB Nvidia Tensor Core GPUs, offering 640GB of GPU memory per node.
The system uses a high-performance, low-latency fabric that enables workload scaling across interconnected clusters, effectively turning multiple instances into one massive unified GPU. Additionally, DGX Cloud is equipped with high-performance storage, providing a complete solution.
The offering will also include Nvidia AI Enterprise, a software tier with over 100 end-to-end AI frameworks and pre-trained models. The software aims to facilitate accelerated data science pipelines and accelerate the development and implementation of production AI.
“DGX Cloud not only provides great computational resources, but also enables data scientists to be more productive and use their resources efficiently,” said Paikeday. “They can start immediately, launch several jobs simultaneously with great visibility, and run multiple AI programs in parallel, with support from Nvidia’s AI experts helping optimize code and customer workloads.”
VentureBeat’s mission it is to be a digital city square for technical decision makers to gain insights into transformative business technology and transactions. Discover our Briefings.