At SIGGRAPH 2018, a professional graphics show, NVIDIA marvelled us when it announced a new GPU architecture capable of Raytracing in real time. Which has been rumoured for months when the first real-time raytracing demo released by NVIDIA and UNREAL ENGINE at the GDC in early 2018.
What Is Ray Tracing?
Raytracing is a rendering method. There are 4 rendering methods. Rasterization (the types used in games), Raycasting, Raytracing and Path Tracing. Raytracing and Path tracing are the types of rendering methods used in Visual Effects. It is very computationally intensive and the results are very physically accurate. Rasterization is not computationally intensive and the results are not physically accurate. That’s why games can run at 60 fps at a low-cost PC because they use the rasterization method. It’s also the reason why in-game graphics using this method will never look as real as that of movies.
The first product that NVIDIA unveiled which uses the new Turing architecture is the Quadro series. This new GPU enables us to render film and TV realism in real-time. Until now, it was extremely costly to even attempt real-time ray tracing, and with the new Quadro RTX series of GPUs, it is achievable
NVIDIA created the new Quadro RTX raytracing GPU based on the new Turing architecture.
What is the Turing Architecture?
Turing combines 4 different processing types in 1 GPU. The first processing type is new real-time raytracing capability using what they call RT Cores. A special part of the circuitry that enables real-time ray tracing for accurate shadowing, reflections, refractions, and global illumination. Ray tracing essentially simulates light, the new Quadros can simulate up to 10 billion rays per second. The second is taking AI deep learning and processing with what they call Tensor cores. The third is using the advanced shading capabilities to draw graphics. And the fourth processing type is for simulations.
NVIDIA’s RTX ray tracing technology was originally shown operating on multiple Tesla V100 GPUs which were the company’s most powerful GPUs at the time.
They also announced Turing for architecture and construction, film and visual effects. This enables the visual effects studios to create photorealistic effects and renders much faster than before. The new RTX technology has already been adopted by some of the world’s biggest visual effects and 3d programs. Such as Adobe Dimension CC, Autodesk Arnold, Clarisse, DaVinci Resolve, Dassault Systemes Catia and Solidworks, Octane Render, ParaView, Redshift, Siemens NX, Unity and Unreal Engine.
The new Turing architecture opens up an estimated market of about $250Billion visual effects industries.
Using NVlink, you can be able to connect 2 GPUs together with speeds of up to 100gb/s
The new Quadro RTX GPUs comes in 3 types.
The RTX 5000 | 16GB/32GB using NVLink | 6 Giga Rays/sec | 3,072 CUDA Cores | 348 Tensor Cores at $2,300
The RTX 6000 | 26GB/48GB using NVLink | 10 Giga Rays/sec |4,608 CUDA Cores |576 Tensor Cores at $6,300
The RTX 8000 | 48GB/92GB using NVLink | 10 Giga Rays/sec |4,608 CUDA Cores |576 Tensor Cores at $10,000.
NVIDIA also announced yet another appliance reference architecture. This new reference architecture from NVIDIA is called the RTX Server and is designed to serve the needs of VFX studios and other companies wanting photorealistic visualizations in near real-time. By stacking eight RTX 8000 GPUs inside dramatically reducing size and cost as compared to using CPUs as you will need more CPUs to get the same amount of power.
Here are sample scene images from videos rendered with this new Turing architecture in real-time.
NVIDIA is crushing new grounds with this leap and opens up doors for the growth of the visual effects studios helping them to do more work in less time with this stunning and powerful chip.