Home / Engineering Resource Center / Get Immersed in GPU-Accelerated Simulation and AI-Driven Rendering

Get Immersed in GPU-Accelerated Simulation and AI-Driven Rendering

The increased computing and graphics processing power in workstations has ushered in a new level of interactivity in conceptual design, usually accomplished in CAD software. In mainstream modeling programs such as SOLIDWORKS, Autodesk Inventor, PTC Creo, and Siemens PLM Software’s NX, users can now instantaneously render what they see in the CAD modeling window to view their design in a photorealistic mode to judge its aesthetic appeal. The near-instantaneous responsiveness allows designers to make quick changes, to compare the design’s effects in different materials or colors, for example. On sufficiently powerful workstations, users may even perform geometry sculpting and modeling operations in a fully rendered mode while maintaining interactivity with the model.

But in simulation and rendering, the compute-intensive nature of the operations has hindered interactivity. Engineers have learned to work around these limitations. One common method, for instance, is to set up the simulation and rendering jobs to kick off during the weekends and after business hours so they won’t interfere with normal design workflows. But the latest developments in the design software sector indicate there are new ways to overcome these bottlenecks in simulation and visualization.

Interactive Simulation

Software-driven simulation, typically done in pixels and bytes in a finite element analysis (FEA) program, doesn’t require the same time-consuming, labor-intensive meticulous setups and cleanups that physical product tests demand. Nevertheless, engineers tend to use digital simulation sparingly, because there’s a cost, calculated in computation. FEA programs put a heavy burden on the computing system, especially on the CPU. It’s not uncommon for some workstations to come to a grinding halt during a particularly complex simulation scenario. Moving, rotating, or modifying the model during an active simulation run would have been ill-advised, because even minute adjustments could trigger recalculations that significantly prolong the run—or crash it, in the worst case scenario.

But the launch of ANSYS Discovery Live, made public in February, brings new possibilities. “ANSYS Discovery Live provides instantaneous 3D simulation, tightly coupled with direct geometry modeling, to enable interactive design exploration and rapid product innovation. It is an interactive experience in which you can manipulate geometry, materials types, or physics inputs, then instantaneously see changes in performance,” the company writes.

The software takes advantage of the parallel processing power in NVIDIA GPUs, effectively enabling real-time visualization of simulation. Built on NVIDIA’s CUDA programming platform, “Discovery Live shortens design-simulation workflows from days and hours to minutes and seconds,” NVIDIA’s Baskar Rajagopalan writes in a blog post.

GPU-powered parallel processing allows instant display of complex fluid flow and thermal results. Image courtesy of ANSYS.

GPU-powered parallel processing allows instant display of complex fluid flow and thermal results. Image courtesy of ANSYS.

The program supports structural, fluid dynamics and thermal simulation. The meshing process, which tends to stump novices and users with limited simulation experience, is automated. It occurs in the background, with little or no user intervention. The simple, straightforward interface and setup is designed to appeal to a much broader user base beyond the usual simulation experts.

ANSYS Discovery Live requires an NVIDIA GPU with 4GB of memory and the latest graphics driver. The recommended choices are NVIDIA Quadro P4000, P5000, P6000, or GP100 from the product line for professional workstations. Workstations equipped with Quadro GPUs and certified for professional use are available from Dell and other leading system providers.

“ANSYS’ Discovery Live unleashes the power of NVIDIA GPUs and CUDA parallel computing to make the first ever real-time design environment with simultaneous visualization and simulation possible,” says Bob Pette, vice president of Professional Visualization, NVIDIA. “We’re committed to working with ANSYS to advance breakthroughs in the way new products are designed and created.”

Beyond the new ANSYS Discovery Live, the use of the GPUs for simulation has shown to be beneficial — especially in computational fluid dynamics (CFD) operations. Due to the amount of parallel calculations possible, the operation lends itself to GPU acceleration. ANSYS Fluent is among the CFD software that supports GPU acceleration.

According to ANSYS: “Integration of AmgX, a library of GPU-accelerated solvers developed by NVIDIA, within Fluent makes this possible. By adding GPUs to existing clusters and workstations, engineers can reduce time to solution by up to half. In addition to speeding up simulation, GPUs consume less energy when compared with a CPU-only solution.”

AI to Speed Up Rendering

Next to simulation, photo-realistic rendering is arguably the most compute-intensive operation in product design. Though once treated as eye candy, photorealistic renderings have now become part of critical design reviews, especially among automotive manufacturers. The rising popularity of augmented reality and virtual reality (AR and VR) devices also indicates the need for real-time, instantaneous rendering will soon become the norm, not an exception to the rule.

Last year, at the GPU Technology Conference (GTC) hosted by NVIDIA, the GPU maker demonstrated a technology that employs artificial intelligence (AI) to speed up rendering. The technology specifically targets the denoising operation, which calculates and resolves light ray bounces and particle positions to create a physically accurate representation of the subject.

In a blog post detailing the technology, NVIDIA states: “Existing algorithms for high-quality denoising consume seconds to minutes per frame, which makes them impractical for interactive applications. By predicting final images from only partly finished results, Iray AI produces accurate, photorealistic models without having to wait for the final image to be rendered.”

Preview of a feature in SOLIDWORKS 2019, showing SOLIDWORKS Visualize rendering an assembly with AI-powered noise removal from NVIDIA. Image courtesy of SOLIDWORKS.

Preview of a feature in SOLIDWORKS 2019, showing SOLIDWORKS Visualize rendering an assembly with AI-powered noise removal from NVIDIA. Image courtesy of SOLIDWORKS.

SOLIDWORKS CAD software’s built-in rendering tool, SOLIDWORKS Visualize, supports GPU acceleration. At this year’s SOLIDWORKS World Conference, NVIDIA demonstrated the new AI-based denoising capabilities of NVIDIA Optix 5.0 on SOLIDWORKS Visualize. SOLIDWORKS users “can render their designs up to 10x faster” with this approach, according to NVIDIA.

“I got a chance to test the pre-release of this Visualize at SolidWorks World 2018 and was extremely impressed”, says Scott Hamilton, Dell Industry Strategist. “I have used many physically based rendering systems over the years and one of the drawbacks has always been the time it takes to completely resolve an image at the pixel level. The AI based denoising solves this problem. The speed up provided by the AI-based denoising dramatically decreases the project cycle.”

See the Denoiser in Action

Check out the “Introducing the New Artificial Intelligence Denoiser” blog post by SOLIDWORKS’ Brian Hillner for more information.

Speeding Up Product Development

In many industries, including technology front-runners like automotive and aerospace manufacturing, the inclination is to perform as much of the simulation and validation in the digital realm as possible. The trend is driven by the dual desire to cut cost and speed up product development in the face of more complex requirements.

“The increased requirements of cars equipped with cameras, sensors, touchscreens, computers and other electronics effectively turns modern cars into large smartphones,” Bjorn Sjodin, vice president of Product Management at COMSOL told DE, for example.

In the aerospace sector, in addition to working with the Federal Aviation Administration to accept more simulations of situations that previously required extensive (and expensive) physical testing, manufacturers are also simulating the effects of new technologies, such as lightweighting via generative design and additive manufacturing.

“We see their combination of generative design, shape optimization and advanced fabrication technologies as absolutely essential to the next generation of more fuel-efficient, more sustainable aircrafts,” says Bastian Schaefer, an innovation manager at Airbus.

Those new shapes need to be simulated in order to be manufactured as efficiently as possible, ensure safety and to save costs associated with physical tests. Therefore, engineers’ reliance on simulation and visualization is expected to increase. GPU-accelerated simulation and AI-driven rendering have much to contribute to the immersive design workflow that leads to shorter design cycles.