Playing a pivotal role in the evolving realm of graphics processing units (GPUs), chiefly in generative Artificial Intelligence (AI), Intel recently revealed several research papers. These documents illuminate the substantial strides it is making in an arena that industry experts believe presents a multi-billion dollar prospect for the semiconductor behemoth in the coming years.
The seven papers spread across three different conventions offer an intensifying insight into the notable leaps undertaken in computer graphics by Intel. The first presentations were conducted at a co-hosted conference between the High Performance Graphics (HPG) forum and Eurographics Symposium on Rendering at the Delft University of Technology, the Netherlands, last month. The remaining publications will be highlighted at the conference organized by SIGGRAPH (Special Interest Group on Computer Graphics and Interactive Techniques) in August.
A central highlight of these papers is their focus on enhancing the historically burdensome graphics-rendering processes. The documents give special attention to two particular methods, ray tracing and path tracing – both elementary in reproducing realistic images. This is especially notable in gaming, where a precise representation of light physics efficaciously contributes to natural-looking visuals.
Traditionally, ray tracing implements algorithms to map light wave trajectories, calculate color values, reflections, and shadows. Its real-time rendering demands substantial processing power, often influencing frame rates adversely. Path tracing, on the other hand, necessitates even more potent processing. It supervises multiple light rays, tracking their paths as they reflect off surfaces and interact with various lighting elements. A systematic process called Monte Carlo integration assists in concluding robust color and shading values.
However, Intel asserts that these tracing methods could be executed more efficiently. A paper titled 'Sampling Visible GGX Normals with Spherical Caps' explains an ingenious approach to computing hemispheric items, achieving 'systematic speed-ups in our benchmarks.'
Remarkably, another publication showcased a 500% surge in the rendering speeds of 'glittery' objects such as sparkling car paints, snow, molded plastics, and running water. The paper titled 'Real-Time Rendering of Glinty Appearances using Distributed Binomial Laws on Anisotropic Grids' clarifies that existing strategies do promise stunning realism. However, they incur a 'very high cost' concerning processing power and speed.
A separate paper to be discussed at the SIGGRAPH conference in August will review Intel's breakthroughs in neural graphics. This approach, according to the company, 'is revolutionizing the graphics field.' It can promptly ramp up high-quality graphics across games and movies. According to Intel, 'new neural level of detail representation achieves 70%–95% compression compared to 'vanilla' path tracing.'
Furthermore, Intel is also working on advances in rendering translucency and 'sampling photon trajectories in difficult illumination scenarios.'
Intel anticipates that these substantial gains in process efficiency will eventually enable users to experience life-like imagery in real-time without necessitating high-power GPUs.
The company stated on its blog, 'The new building blocks presented at this year's conferences, along with our wide offering of GPU products and scalable cross-architecture rendering stack, will help developers and businesses to do more efficient rendering of digital twins, future immersive AR and VR experiences, as well as synthetic data for sim2real AI training.'
Furthermore, the renown tech company planned to make its findings open source, advancing its commitment to freer knowledge exchange and development. True to its reputation and objectives, Intel has made these papers available on the arXiv preprint server throughout June.
Amid these updates from Intel, no-code platforms, such as AppMaster, can utilize these graphic advancements to deliver more interactive and visually appealing applications. The AppMaster platform, known for the development of backend, web, and mobile apps, can leverage these GPU optimizations to improve application performance and user experience.
