The word Ray Tracing was unknown to personal computer users, since it was a word more related to the world of professional 3D animation. Since talking about Ray Tracing was talking about complex systems made up of several computers in parallel, occupying an entire room. Well, the RTX 40 have a secret inside that will cause two things in the future. The first is to further differentiate itself from AMD in this aspect in terms of performance, the second is that it will make the adoption of ray tracing in games even easier.
Soon it will be five years since the launch of the RTX 20, and the launch of a commercial game with only Ray Tracing has not happened. So a certain NVIDIA prediction has not come true. If we look at it cold, the same time has passed between the RTX 20 and the RTX 40 as between the first Voodoo Graphics and the original GeForce. In other words, the evolution in terms of graphic architectures is less today than at the end of the last century.
What is the secret of the NVIDIA RTX 40?
Since we are going to comment on a patent and these have as their objective the resolution of an existing problem, the first thing we have to do is explain what that problem is. The most basic idea of the Ray Tracing algorithm is that for each pixel in the image a ray is fired from the viewer’s perspective, and then it is determined whether there is an object in that pixel or not, and if there is, it is painted a color. determined and so on.
However, its visual advantage lies in the fact that we can also summon lightning from objects and this is where the big problem comes in. If there is no boundary condition, the rays bounce around the scene indefinitely and the complexity of the algorithm grows to infinity, not being able to run and freezing the computer.
So you have to put a series of completion conditions:
- The beam exits the scene into the void, so its corresponding instance is removed.
- A certain number of bounces per ray is assigned, once reached that amount the ray is destroyed.
- Each object is given a refractive quotient ranging from 0 to 1, a 1 is a mirror material, a 0 is literally a black hole. When said ratio in a ray is too low, then it is destroyed.
- Determining the distance between two objects, if it is too great, then the beam is destroyed.
Well, the secret of the RTX 40 has to do precisely with the last point.
The application on current graphics cards.
Once we have specified the problem we have to talk about the solution. What happens is that in the case of the Vulkan and DirectX APIs, used for games. In them the rays are not executed automatically, but have to be explicitly called by the game code. Thus, we can suddenly find scenarios where Ray Tracing disappears totally or partially in order to maintain a high frame rate.
The secret of the RT Core of the RTX 40
Well, in the third generation RT Core and if we pay attention to a certain patent published in the last year by NVIDIA, there is a good chance that said unit will receive a significant improvement. These would be two new units that will serve to speed up ray tracing.
- The first one is capable of determining the distance between the camera and an object or between two objects . Which saves power to the rest of the graphics chip in this task. So if the second object is too far away from the first one from which the beam is invoked, then the beam that would go from one unit to another is marked as disposable Another thing you can do is decrease the power of the light source in order to improve the representation of shadows.
- As for the second unit, it would be in charge of determining if it invokes more lightning according to the conditions of the scene . That is, those that have been discarded due to the effect of distance, can then be recovered in other objects. So in the end everything is based on an intelligent use of the resources that we have available.
We have not invented all this, but it belongs to the information of an NVIDIA patent entitled ADAPTIVE RAY TRACING SUITABLE FOR SHADOW RENDERING from where the entire process is described and from where we have taken the image below.
NVIDIA’s goal with the RTX 40 is to leave out AMD.
NVIDIA’s idea would be none other than to automate Ray Tracing and, therefore, make all games compatible without the need for patches of any kind by programmers. This would give them a huge advantage over AMD, since this technique could only be executed on their graphics cards.
If we add to this that from one frame to another the vast majority of objects in the scene do not vary their distance between them since they are static, then the system has it much easier to assign which rays are invoked and which are not. In any case, the ultimate goal is for games to get higher frame rates on NVIDIA cards compared to their direct rivals.
By the way, although this is less reliable, it is said that NVIDIA could increase the ratio of RT Cores from unit per SM to unit per subcore, that is, quadruple their number compared to the current RTX 30 . Although take this rather with tweezers, in any case, if so. It would be an important punch on the table against its rivals and one more step towards the massive adoption of Ray Tracing in games.
