oneAPI is Accelerating Graphics-Intensive Workloads Across Industries

December 1, 2022

oneAPI has been expanding into enterprise and consumer industries like film, animation and gaming, design and architecture, automotive, as well as the medical industry. The specification is helping create high-performance, high-fidelity visual experiences with Intel oneAPI architecture. 

The graphics processing unit, or GPU, is one of the most important types of computing technology, both for consumers and businesses. Designed for parallel processing, the GPU is used in a wide range of applications, but its explosion in popularity is due to graphics rendering and ray tracing. 

oneAPI is Powering the Future of Gaming 

A new era of breakthrough performance and efficiency in graphics workloads has been unfolding. This has been particularly advantageous for the world of 3D graphics, which has rapidly expanded into the domain of mobile platforms like cell phones. 

The oneAPI Rendering Toolkit is a collection of libraries that includes Embree, Open VKL, Open Image Denoise, and Open Path Guiding Library. These libraries are part of Intel’s oneAPI suite focused on rendering, which is most common in video games or interactive graphics where the 3D images are calculated at a very high speed so that it looks like the scenes, which consist of multitudes of images, occur in real time when players interact with the game.

Embree is an open-source library that solves the fundamental computations for ray tracing- based rendering. It uses the latest ray tracing algorithms, and kernels are highly optimized to deliver a 1.5 to 6x speedup. The library makes high-performance, ray-based visibility queries available, and it provides the spatial index structures over the primitives of a 3D scene, as well as the capability to quickly build high quality data structures needed to perform the visibility queries operations. This is achieved by using SIMD (AVX-512) and optimized data structures, among other optimizations. The library is meant to be utilized in professional rendering applications and has found wide adoption in filmmaking. In gaming, it is used for precomputing lighting for static geometry, such as in Activision’s Lightmapper. It can also be used for ray-based collision detection, including gun shots. 

Intel’s Open Volume Kernel Library (Open VKL) is a collection of high-performance kernels for sampling and traversing rays in volumetric data (scalar fields). It contains APIs for single sampling and packets to aid in vectorization of ray tracing algorithms. It is optimized for x86 CPUs and includes AVX-512 support. Open VKL vastly improves volume rendering applications used to display a 2D projection of a 3D discretely sampled data set, typically a 3D scalar field, using a wide variety of volumetric data structures, as well as performance-optimized traversal and sampling algorithms. 

Open Image Denoise is a denoising library for images rendered with ray tracing. It provides a high-quality deep learning based denoising filter, and it enhances interactive preview and final frame rendering. The library can run on any modern CPU and also supports up to AVX-512. Open Image Denoise can be used for denoising light maps in Unity, for example, which gives improved quality over Gaussian filters with a reduced number of samples, and it is independent of specific GPU hardware vendors.  

Open Path Guiding Library (Open PGL) is the industry’s first open-source library that enables developers to easily integrate state-of-the-art path-guiding methods into their renderers.

Open PGL increases rendering performance by optimizing the exploration of complex light transport, leading to a substantial reduction in image noise generation. This noise reduction allows artists and developers to use more realistic lighting when building and designing scenes and animations, while still offering artistic freedom at the same or better rendering performance.

These libraries have become industry-trusted building blocks, which are integrated into many third-party professional content creation tools and applications for broader ecosystem use, helping creators and engineers push the boundaries of visualization to develop amazing, immersive visual effects and industrial visualizations.

oneAPI is Revolutionizing Graphics in Film and Animation

Although digital animation and special effects developed for movies are cheaper and faster than traditional 2D-cel and stop-motion animation, they require highly intense compute workloads. Computer animation can be either 2D or 3D, but 2D animation often involves virtualization of the traditional 2D animation workspace. Either method requires multiple groups of networked computers to render scenes. The oneAPI Rendering Toolkit is enabling ways to reduce the enormous number of resources and long production timelines required to produce animated scenes and visual effects. 

The toolkit libraries are helping movie studio creators produce more vivid photorealistic visuals in less time than with traditional methods and with reduced computer processing. Cinesite, for example, a well-respected VFX, animation and digital entertainment studio, was using Arnold, a popular render software made by Autodesk, to render images for the feature film Addams Family 2. Intel worked with Autodesk to integrate Intel Open Image Denoise into Arnold. The oneAPI Rendering Toolkit library was used to remove noise from the imagery with its collection of machine-learning algorithms and efficient deep-learning-based denoising filters, which are trained to handle a wide range of samples per pixel. The library not only helped to create sharper, more detailed visual effects but also helped the studio gain “10% to 20%—and sometimes 25%—efficiency in rendering, saving thousands of hours in rendering production time,” according to Kenny Chang, head of Lighting at Cinesite. 

With use cases like Addams Family 2, there’s little doubt that utilization of oneAPI’s rendering and ray tracing libraries will dramatically expand across Hollywood studios, ushering in accelerated workloads with large data sets and high complexity to create high-performance, high-fidelity visual experiences the likes of which have never been seen before. 

In 2021, the Embree library received an “Oscar” in the Scientific and Technical Achievement category from the Academy of Motion Picture Arts and Sciences in recognition of its role as a contributing innovation in the movie making process.

oneAPI is Boosting Digital Design and Architectural Visualization Capabilities 

In creative business sectors like architecture, interior design and product design, the oneAPI Rendering Toolkit is helping companies render computer graphics for 2D and 3D design conceptualizations of planned physical spaces. 

Integrated with professional creator applications, the toolkit enables creators to develop 4K quality rendered graphics by providing them the computational horsepower needed to manage large data and rendering tasks. A creator can digitally design a modern home, for example, with structural details specific to its real-life construction, coupled with furniture and accents to help customers visualize every aspect of the scene—from wire frames to illuminated shading—using a film quality path tracer built with the Embree ray tracing library. Being able to create photorealistic, 3D visuals, creators and companies can build digital construction scenes that are interactive, immersive, real-time experiences to help potential customers, vendors or contractors fully imagine a close-to-final construction project before one brick is ever laid in real life, saving businesses both time and costs. 

The oneAPI Rendering Toolkit essentially allows creators and their companies to give their clients a real-time, screen captured, photo realistic walk-through of a planned construction project. By being able to accurately and strikingly convey graphics that contain the highest fidelity, complex lighting, and unique look and feel of a planned construction project, creators are now empowered to exponentially improve the creative design process. 

Talcik & Demovicova, for example, is a digital boutique studio specializing in high-end architectural visualization; they create renderings for clients in the architecture, interior design, product design, and real estate sectors. The company has been in business for over a decade, and it has evolved to combine traditional design methods with computer graphics using the oneAPI Rendering Toolkit to optimize Corona—a high-performance photorealistic renderer— and reduce memory displacement. 

OSPRay Studio helps Bentley Motors address their most demanding visual workloads

OSPRay Studio is an open source, interactive visualization and ray tracing application that leverages OSPRay as its core rendering engine. It can be used to load complex scenes requiring high fidelity rendering or very large scenes requiring supercomputing resources.

OSPRay is an open source, scalable, and portable ray tracing engine for high-performance, high-fidelity visualization on Intel Architecture CPUs. It provides an open, powerful, and easy-to-use rendering library that allows developers to easily build applications that use ray tracing-based rendering for interactive applications, including both surface- and volume-based visualizations. OSPRay is completely CPU-based and runs on anything from laptops to workstations, to compute nodes in HPC systems. 

Using OSPRay Studio and other components of the oneAPI Rendering Toolkit, Bentley Motors—one of the world’s most sought-after luxury car brands for more than 100 years—created a virtual, interactive, 3D vehicle showroom modeled after their CW1 flagship showroom located in Crewe, England. In the words of a Bentley representative, “OSPRay is able to render them [cars] beautifully.” Using a computational infrastructure consisting of 10 Intel xeon server nodes with Intel optane persistent memory—using no GPUs or accelerators—Intel engineers collaborated with Bentley to create more than just a virtual showroom; it’s a digital platform housing 12 highly detailed 3D vehicles that customers from around the world can configure to fit their personalization desires. Each virtual vehicle contains over 20 million primitives each, and the full scene, showroom and props totals over 315 million triangles and textures occupying 120 gigabytes of memory. 

oneAPI is Helping to Innovate Processing Systems for Medical Imaging 

Medical imaging is the field of medicine where devices like X-ray, ultrasound, CT scan and MRI machines are used to create 2D and 3D visual representations of the interior of the body such as organs and tissues for clinical analysis and medical intervention. 

OSPRay is also being used for visualizing human-scale blood flow simulation. Intel recently partnered with the Leibniz Supercomputing Centre (LRZ) and University College London to exhibit a demonstration visualizing blood flow of the forearm of a virtual human. CompBioMed—a European Commission H2020 funded Centre of Excellence focused on the use and development of computational methods in the biomedical domain—is creating a virtual human: a personalized digital representation of a human’s biophysical processes. A component of this process involves HemeLB, a highly scalable, 3D fluid dynamics code that is developing high-resolution simulations of blood flow throughout human-scale vasculature. Visualizing the results, however, is a significant challenge because of the complexity and volume of the data. 

OSPRay Studio is used to visualize the resulting data directly on the production machine, LRZ’s supercomputer, SuperMUC-NG, and producing outputs from 2D to virtual reality. A custom plugin was created to map the data to the highly memory efficient vector-valued volumes (VDB) volumes of Open VKL for delivering visualization at native resolution and interactive frame rates, eliminating data pre-processing. 

Another example involves the S-Fetus 4.0 Obstetric Screening Assistant utilized by doctors to perform obstetric sonography, or prenatal ultrasound, in an effort to reduce maternal and perinatal mortality. Because traditional obstetric screening methods require a lot of time and manual operations done by medical professionals, SonoScape—a medical manufacturer of trolley and portable ultrasound systems and high-definition endoscopy systems—launched a smart obstetric screening system. This system automates much of the results data generated by the sonography screening process, and it is powered by the oneAPI Base Toolkit. Because results data is automated, it reduces the workload of doctors. 

Using the Intel VTune Profiler, SonoScape was able to boost the computation performance of the obstetric screening assistant by identifying areas where they could streamline CPU and GPU workloads utilizing vector and scalar processing. SonoScape then replaced the areas of lagging computational performance with Intel Integrated Performance Primitives (Intel IPP) APIs to accelerate image processing, data compression, and application performance, among other functions. Intel IPP is a cross-platform software library of functions for multimedia and data processing applications. SonoScape also utilized the Intel DPC++ Compatibility tool to migrate existing CUDA code to DPC++ more efficiently to ensure cross-architecture compatibility. The integration of oneAPI architecture with S-Fetus 4.0 Obstetric Screening Assistant AI technology resulted in faster computational performance, heterogenous scalability, cost-savings—and best of all, enhanced patient care.