Quadro RTX Powers Visual Computing Technologies

A new series of GPUs includes features designed specifically for real-time ray tracing, virtual reality, engineering simulation, and artificial intelligence.

In an increasingly competitive marketplace, manufacturing and design teams have to grapple with a number of challenges, such as time to market, cost reduction and unique product innovation. To meet these challenges, a range of new technologies are converging to change the way products are created altogether. These include developments such as real-time ray tracing, virtual reality (VR), real-time engineering simulation, and artificial intelligence (AI).

OpenGL rendering (top), similar to what is viewed in typical CAD viewports compared with RTX ray tracing (bottom).(SOLIDWORKS Visualize technical preview; Model courtesy of GrabCAD.)

OpenGL rendering (top), similar to what is viewed in typical CAD viewports compared with RTX ray tracing (bottom).(SOLIDWORKS Visualize technical preview; Model courtesy of GrabCAD.)

While these technologies are becoming increasingly accessible for a variety of reasons (dropping costs, hardware improvements, etc.), they also require much more computational power. This means that, if a business wants to use VR, for instance, specialty workstations may be provided to a given expert within an organization, such as a visualization or VR specialist, keeping the technology just out of reach for most design teams.

Fortunately, high-powered computing is becoming accessible as well, due to increasingly powerful graphics processing units (GPUs). Most recently, NVIDIA released its Quadro RTX GPUs, which are based on the new Turing architecture. The Quadro RTX series is said to deliver more computer power for real-time ray tracing, VR, real-time simulation and AI capabilities.

Real-Time Ray Tracing

Ray tracing is a rendering technique in computer graphics used to generate photorealistic images by simulating the behavior of light in a scene. Quadro RTX GPUs have been designed with ray tracing in mind, introducing the ability to perform this rendering process in real time.

The Turing architecture of the Quadro RTX series goes beyond the previously released Pascal architecture by adding new types of cores. In addition to the fundamental CUDA Cores long found in NVIDIA GPUs, the Quadro RTX features Tensor Cores, dedicated to deep learning and AI applications, as well as the entirely new RT Cores.

The Quadro RTX 4000 featuring 8GB of GDDR6 memory, 36 RT Cores, 288 Tensor Cores and 2,304 CUDA Cores makes it possible to perform real-time ray tracing. (Image courtesy of NVIDIA.)

The Quadro RTX 4000 featuring 8GB of GDDR6 memory, 36 RT Cores, 288 Tensor Cores and 2,304 CUDA Cores makes it possible to perform real-time ray tracing. (Image courtesy of NVIDIA.)

For professional visualization and advanced product design workflows, the RT Cores in the GPU are able to perform specific ray tracing calculations faster than CUDA Cores. This speeds up ray tracing while freeing up CUDA Cores to work on other tasks. Designers and engineers who previously have had to wait for scenes to render, or re-render, every time lighting was moved or design modification was made, now have the ability to see photorealistic scenes in a dynamic environment.

In addition to using RT Cores to improve rendering time, Tensor Cores enable the use of AI to speed up the rendering of photorealistic images, with trained neural networks removing the “noise” that is visible while the image is being rendered.

Andrew Rink, head of Marketing Strategy for AEC and Manufacturing Industries at NVIDIA, described how this process works: “AI is used to train a neural network with millions of rendered images so that it can recognize noise in an image as it is being rendered. The neural network can then use its training and information about the surrounding pixels to intelligently fill in the missing pixels, significantly speeding up the time it takes to create the full image. When you rotate your model, the AI kicks in so that it’s instantly resolving that render—any changes in those pixels because of the light, the angle, the color, the reflection—very quickly. And then, after a certain number of passes, the regular renderer starts to kick in.”

An array of monitors being coordinated with NVIDIA Quadro Sync, made possible with the RTX line. (Image courtesy of NVIDIA.)

An array of monitors being coordinated with NVIDIA Quadro Sync, made possible with the RTX line. (Image courtesy of NVIDIA.)

To get a sense of how real-time ray tracing might impact the design process, imagine a product engineer designing a coffee pot, with SOLIDWORKS on one display and SOLIDWORKS Visualize on another. Every time the model is changed in CAD, it is instantly updated and rendered photorealistically on the other screen. The CAD model can be rotated or viewed from a different angle while maintaining the lifelike rendering of the coffee pot, with light passing through glass or reflecting off chrome, just as it would in the physical world. Assessing digital models in such a predictable manner helps designers make faster and more confident decisions about materials and aesthetics while they are designing.

Truly Immersive Virtual Reality

The advantages of real-time ray tracing naturally carry over to other emerging technologies like VR. The price of head-mounted displays (HMD) is dropping dramatically, but great VR experiences require serious amounts of computation.

GPUs enable quicker rendering of VR images so that design teams can make better-informed design decisions and identify flaws earlier in the design process. As a result, more design work can be performed virtually before investing in the cost of physical prototypes. And, as a product heads to market, marketing and sales teams can create VR assets for promotional purposes.

VR is still in the relatively early stages of adoption; however, the Quadro RTX line has been designed to be powerful enough for even future iterations of VR technologies. Additionally, the new Quadro RTX GPUs all include a feature called VirtualLink, an output port for communicating with HMDs through a single USB-C interface. This reduces the number of cables required and results in a less tangled VR experience.

Real-Time Engineering Simulation

While real-time ray tracing can provide a photorealistic visualization of a product and VR can add new depth to that visualization, engineering simulation can help ensure that a product actually behaves in the physical world as a designer wants it to.

Engineering simulation is a computationally intensive and often arduous process, but, last year, ANSYS made a breakthrough with ANSYS Discovery Live. For the first time, the company made it possible to perform simulation without the need for complex computer-aided engineering software run by simulation experts.

A screenshot of ANSYS Discovery Live. (Image courtesy of ANSYS.)

A screenshot of ANSYS Discovery Live. (Image courtesy of ANSYS.)

Early in the design process, a tool like Discovery Live can help users to evaluate different design options before committing to a final design, as well as allow them to achieve to optimize designs. A flange could be strengthened if it is found to be weak. A bike helmet could be reshaped if it results in too much drag. An aerospace component could have unnecessary material removed to reduce weight. Changes can be made before it becomes too expensive to make them, resulting in time and money saved.

This real-time simulation can only be performed with the use of NVIDIA GPUs. In particular, the solvers in Discovery Live take advantage of the parallel processing in Quadro RTX GPUs. Though the capabilities of NVIDIA cards aren’t always fully exploited by software developers, ANSYS’s new software leverages their power to speed up the design process.

As a result, all engineers can use the software, even without a background in computer-aided engineering. Much of the functionality of this software has since been incorporated into PTC Creo with Creo Simulation Live, thus extending the tool to hundreds of thousands more engineers. 

Artificial Intelligence in Design

Perhaps the most intriguing emerging technology to assist in the design process is AI. Though the technology is finding itself in a variety of design applications (see real-time ray tracing above), one particularly interesting way in which it is impacting how items are made is through generative design.

A generative design process starts with a goal (such as a specific weight), along with forces and constraints on a shape. The software, trained on numerous previous designs, uses simulation to determine where material is needed and where it isn’t. There can be an infinite number of shapes that satisfy the initial conditions, but a generative design will typically stop after a few hundred. A design exploration (usually a visual inspection) can consider other criteria, such as manufacturability or even aesthetics.

Having so many simulations in the course of running a generative design model make the process computationally intensive. GPUs are NVIDIA’s answer to speeding up this process. The Tensor Cores, in particular, are designed to handle the deep learning necessary to train generative design software.

With a workstation that uses a powerful GPU designed for AI applications, engineers can begin performing generative design at their desks, whether it be a 3D-printed seatbelt bracket with an optimized topology or an organically inspired piece of home decor.

Quadro RTX-Accelerated Workstations

One can imagine a workflow that utilizes all of these technologies becoming a norm to speed up the design process, reduce costs and introduce innovative features to a product. Real-time ray tracing can allow a designer to better visualize their work before using VR to get a virtual sense of the product’s presence in the physical world. The design’s physical performance can be perfected using real-time engineering simulation and AI tools like generative design can ensure that the end product looks like nothing anyone’s ever seen before. While photorealistic, immersive design and engineering simulation can streamline the design process before costly mistakes are made later on, AI-enabled generative design can enhance the product’s performance in the physical world. 

All of this is accelerated or enabled with the use of a quality GPU. Since powerful GPUs have become accessible to designers and engineers, it’s no longer necessary to find a supercomputer to take advantage of these new design technologies. Workstations equipped with NVIDIA Quadro GPUs are more powerful than previous generations—they are even capable of running the entire design and engineering workflow from end to end on a single workstation.

The NVIDIA Turing GPU architecture powers the new generation of GPUs. (Image courtesy of NVIDIA.)

The NVIDIA Turing GPU architecture powers the new generation of GPUs. (Image courtesy of NVIDIA.)

The baseline RTX 4000 ($900) includes 8GB of GDDR6 memory with up to 416GB/s of memory bandwidth and 2,304 CUDA Cores; 288 multi-precision Tensor Cores for up to 57 TFLOPS of deep learning performance; and 36 RT Cores, which provide up to 6 Giga Rays per second of real-time ray tracing. The series climbs all the way up to the RTX 8000 with 48GB memory, 72 RT Cores and 576 Tensor Cores.

While the most powerful GPU probably isn’t required for most tasks, it may be necessary for CAD users who are running simulations with millions of elements or those who are embracing photorealistic visualization in a dynamic scene. For most engineers, however, the RTX 4000 or 5000 will provide plenty of power.

Regardless of which GPU one chooses, the Quadro RTX series brings technologies like real-time ray tracing, VR, real-time engineering simulation and AI to designers and engineers in a convenient single-slot form factor that fits in a variety of workstation chassis.

To learn more about the technology, visit the NVIDIA website.

NVIDIA has sponsored this post.