In this year’s GDC event, NVIDIA showed a latest version of its Zora Demo from the Jeffers RTX 50 Series launch, including changes that include control that allows someone to see features like RTX mega geometry. In the panel discussion about the demo and everything, NVIDIA confirmed one thing that we all doubt, but one thing that could surprise you – Zora AH is 100 % -ray tracing and thus it is faster if it used rustyization.
As long as I wander around with 3D graphics (about 30 years), I like to see GPU shopkeepers issuing a cool stand demo to display some of the features of new rendering trucks or fancy hardware. Throughout the years, they have fallen to some extent, with their place to showcase your new graphics card.
This does not mean that NVIDIA’s power demo is not vigorous-when it is seen in real time on a large OLED monitor, it is quite surprising-but what can be done by moving large-scale round posts on the day of the demo. The problem with Zora is not that it is bad, just that the closest reference (such as the game that uses the path to track) is equally impressive.
To discuss its recent progress in GPU technology and graphics rendering, the NVDIA held a panel discussion at GDC 2025, and one of the panelists dropped a slight piece, which was focusing on me which was being talked about (Jet Jait). John SpatzerNavalia’s developer and Performance Technology VP, was willing to do: Zahara Demo:
“There is no ritualization here. All this is including primary rays, including rays. The surprising thing is that it is in fact faster than rising them, so it is not done because it is good to say that in the demo. In fact, in this case, the right thing is to do.”
Given that Zora is an exhibition of NVIDIA RTX neural rendering technologies as well as a complete suit of RTX mega geometry, the fact is that it is a fully -developed demo. However, the fact is that making the same scenario through traditional techniques is faster that it may indicate the point that we now have hardware so that it is able to fully abandon the traditional recurrence.
Well, not right now, because Zora Demo is not the fastest, even an RTX 5090, and it needs every performance trick that can be taken on the RTX neural rendering, RTX mega geometry, and DLSS4 table. That said, this shows the benefits that AI rendering offers and when we are aware that only being used for advanced and frame generation, in the coming years we will see that it is being exploited rapidly to make the fanish graphics at the game worth the frame rate.
And it all does it through the strength. For example, the RTX nervous content takes a lot of detailed materials long, complex shaders, and uses a small nerve network that represents how the light interacts with the substance that results in a good idea of ​​the original thing, except that it all works fast. In this way, as AI can be used to interrupt the entire frame, it can now be used to interrupt a particular shader.
The RTX neural radiation cache does the same thing once or twice in a scene. As a result of the application of a nervous network, it does the same thing and then estimates what the final result of hundreds or thousands of bounces will be.
Indeed, as all the closest, all these AI offering are not perfect and the argument can never happen, but there are neither the techniques of rusticization that we are all aware of. It just needs to be good that you can’t tell during gaming.
It is possible to have a full -wipped scene on the game capable frame rate, because of the power of AI, which follows the set algorithm and physical laws.
And this is not the only NVIDIA that is going down the road: AMD and Intel did not include the Matrix Core in their GPUs for upscalling or frame general. Radling is always playing a nearby, of course,, so, so it’s just an evolution of how we turn the shining dots on the screen into worlds that assure us that this is the fact.
Like it or not, AI is really the future of graphics.
