Main content

NVIDIA paves more roads to USD

George Lawton Profile picture for user George Lawton March 22, 2023
The first despatch from NVIDIA's GTC conference.


The two big themes at NVIDIA’s big GTC conference this month were generative AI and the Industrial Metaverse. The generative aspect tends to be attracting the most hype lately, thanks to consumer buzz around OpenAI’s ChatGPT. But enterprises will need better ways to digitize their physical assets to apply generative AI more broadly. That’s where the Universal Scene Description (USD) language comes in.

The USD format on its own just describes the arrangement of 3D objects and their geometry. But NVIDIA has been building the ecosystems of connections, experiences, and enhancements to turn USD into the HTML of the Metaverse. NVIDIA is making sure that all of this runs better on its own kit in the cloud, enterprise data centers, or high-end laptops. Over 300,000 users have downloaded Nvidia Omniverse tools for developing USD-powered apps. But NVIDIA is also surfacing access to the Omniverse across a rapidly growing ecosystem.

Last year, the Consumer Metaverse drove the most press. This year, the Industrial Metaverse may drive the most profits thanks to USD enhancements, partnerships, and integrations. Richard Kerris, VP of Omniverse Platform Development at NVIDIA, said:

“The Industrial Metaverse is where the businesses are revolutionizing their workflow, so we will go deep on automotive workflows and others in that digital twin space. There is so much we have to offer on the platform and tools that they are using today. I think some of those we will see consumer apps come out of that. Really our initial focus is around the Industrial Metaverse, those 3D workflows all the way through to the operation of a digital twin.”

At the conference, the graphics leader announced better USD connectivity to various industrial apps, 3D collaborative experiences in Microsoft Teams, and SimReady, a new format for exchanging simulation assets on top of USD. These developments will make it easier to share digital twins across tools, collaborate on problems, simulate variations, and generate new designs, processes, and experiences.

In addition, NVIDIA is partnering with Microsoft on the first Omniverse Cloud PaaS service to be rolled out later this year. This new offering will support tools for creating digital twins and publishing 3D experiences. It will also help streamline the process of training robots and autonomous vehicles and generating synthetic data for computer vision apps.

In the official announcement, Nvidia CEO Jensen Huang said:

Every manufactured object, from massive physical facilities to handheld consumer goods, will someday have a digital twin, created to build, operate and optimize the object. NVIDIA Omniverse Cloud is the digital-to-physical operating system for industrial digitalization, arriving just in time for the trillions of dollars of new EV, battery and chip factories that are being built.

Only time will tell if digitalization enters the lexicon as a new term for creating digital twins as opposed to just digitizing the world.

New connectors

First up, NVIDIA has rolled out a bunch of new Omniverse Connectors that promise to improve USD integration across apps from different vendors. New connectors are immediately available for industrial automation and design tools from Siemens, Blender, Cesium, Rockwell Automation, Unity, and VectorWorks. In addition, new connectors are soon expected for tools from Azure Digital Twin,, FlexSim and NavVis Omniverse Connectors are coming soon. In addition, Bentley launched the commercial availability of LumenRT, the first engineering software platform built on Nvidia Omniverse, which also connects to Bentley’s own iTwin platform.

Each of these vendors is not just providing new software but also providing an onramp to help enterprises consume more NVIDIA kit. However, there is still more work to be done to smooth out the limitations of USD, which cannot directly support many of the underlying models in industrial applications. For example, Jupiter Tessellation (JT) rendering is an open standard for exchanging 3D data for product visualization, collaboration and digital mockups developed by Siemens. It supports product structure, lighting, textures, precise geometry, and product manufacturing information.

Noah Cole, senior director of global communications of Siemens Digital Industries Software, said:

Through the newly released connectors, Siemens and Nvidia are making it easier for our customers to experience in high-fidelity and with photo-realism the digital twins created with Siemens Xcelerator within Omniverse. Once imported into Omniverse using the new connectors, these digital twins can be combined with 3D assets created in other tools for a more contextualized visualization. Siemens continues to support USD, and we are also in discussions with Nvidia about adding native JT rendering, which would eliminate some of the process steps that are required in the current USD pipeline.

Improved experiences

Earlier this year, I heard some early adopters in the construction field describing the differences in workflows that ran across bespoke integrations into rendering engines compared to native USD integrations. At the time, they found that the customized connections created a much smoother experience for showing construction project ideas to clients. However, it was also cumbersome to make changes as they had to return to the original tool to ensure the change propagated. In contrast, they found that direct Omniverse integration allowed them to make changes in any tool that were automatically propagated across all the other ones. However, the rendering experience was a bit slower.

NVIDIA’s new hardware and cloud service offerings rolling out this year promise to improve the user experience while also streamlining workflows across tools. Bentley believes the combination of better integration and faster hardware will streamline workflows across partners in larger infrastructure projects. Lori Hufford, Bentley Systems’ VP of Engineering Collaboration, said:

The combination of Bentley iTwin and NVIDIA Omniverse provides a high-performance user experience for infrastructure organizations at a scale that has previously not been possible. The results achieved by the early adopter users of LumenRT for NVIDIA Omniverse demonstrate a unique advantage for project teams working with infrastructure digital twins by providing powerful new digital twin-native workflows that make visualization advantageous to apply on projects of every scale.

My take

The path to digital twins is still a work in progress, impeded by a wide range of processes and file formats used across tools. Modern web apps were driven by enhancements across back-end databases, messaging formats like JSON, and rendering languages like JavaScript. USD on its own is not very capable, in the same way that the HTML specification only supported static web pages initially.

Today, USD is in the same place that HTML was in the early 1990s. NVIDIA has taken on the arduous task of helping to fill in the gaps and improve tools across various leaders. Early adopters are giving impressive demonstrations of what might be possible with improved digital twin interoperability. Today, enterprises must cobble together these solutions on their own.

The native cloud support later this year might bring a little momentum to USD as the universal glue for the Industrial Metaverse. It could also plant the seeds for new tools to generate digital twins. However, more work will be required to standardize ways of describing how things function rather than just how they look. SimReady is a promising step.

A grey colored placeholder image