During a packed morning session at the company's GPU Technology Conference (GTC) in San Jose, Nvidia went into detail about the company's recently-launched RAPIDS software platform. RAPIDS aims to drive the use of Nvidia's GPUs, already widely used for AI/deep learning and high-performance computing (HPC) workloads, to handle traditional enterprise analytics and machine learning workloads, which have historically been handled by server CPUs (quite often Intel's (INTC - Get Report) ).
Examples of such workloads include detecting credit-card fraud and forecasting inventory trends. In a hands-on demo conducted during the session, RAPIDS -- running on Alphabet/Google (GOOGL - Get Report) cloud computing instances featuring Nvidia's Tesla T4 GPUs -- was used to analyze a retailer's Black Friday sales. With a limited amount of work, developers could use RAPIDS to help figure out which products sold by the retailer fit into multiple product categories, and how such items sold on Black Friday.
Nvidia stressed that RAPIDS, which relies on the popular Python programming language, is accessible to corporate data scientists, rather than being the purview of AI researchers. The company also highlighted the platform's support among notable open-source developers, as well as a roadmap in which Nvidia plans to have RAPIDS support a wide variety of tasks related to data science and machine learning, as well as complex graph analytics jobs that attempt to derive insights from the data provided by a set of interconnected nodes (for example, nodes on a financial or social network).
For data analytics workloads, Nvidia is aiming for RAPIDS to ultimately deliver performance improvements of up to 5x to 15x relative to conventional, CPU-based approaches. For machine learning workloads, it's aiming for improvements of up to 10x to 20x. And for graph analysis, it's aiming for improvements of up to 100x to 500x.
Separately, a pair of engineers from open-source software giant Red Hat (RHT) , which is set to be acquired by IBM (IBM - Get Report) , went over their company's work with Nvidia to optimize Red Hat's software for Nvidia's GPUs and hardware.
Red Hat has certified the use of the Red Hat Enterprise Linux (RHEL) operating system on Nvidia's DGX-1 and DGX-2 servers -- they're packed with Nvidia GPUs, and are used for AI and HPC workloads -- and Nvidia has created driver packages to help companies rapidly deploy GPUs on systems running Red Hat's software.
In addition, the companies have worked to create "tuned" software profiles for DGX systems running RHEL that optimize the system's performance for various workloads. And they've collaborated on GPU support for Red Hat's OpenShift software platform, which is used to manage large clusters of app containers. Containers are an alternative to the server virtual machines that have exploded in popularity over the last several years thanks to their portability, resource efficiency and the speed at which they can be deployed.
Collaborations such as the one Nvidia has in place with major software developers such as Red Hat are a significant competitive strength as its server GPU business deals with intensifying competition from larger chip developers such as Intel and AMD (AMD - Get Report) , as well as from a slew of startups.
Nvidia shares fell 0.5% to $168.95 on Monday but are up more than 26% this year.
Will I Have Enough Money to Retire?
Want to learn about retirement planning from some of the nation's top experts? Join TheStreet's Robert "Mr. Retirement" Powell live in New York on April 6 for our Retirement Strategies Symposium. For a limited time, tickets are available for $99 for this full-day event. Check out the agenda, learn about the speakers and sign up here.