Nvidia Zorah: UE5 RTX TechDemo Available as a 108GB Download
Nvidia Zorah: UE5 RTX TechDemo Available as a 108GB Download
At GDC 2025 in March, Nvidia announced the “Neural Shader,” which was also newly introduced with the Blackwell AKA RTX 5000, which is also coming to DirectX this month. The demo is now available as a large 108GB download. An RTX 5000 and Unreal Engine 5 are required for its use.
Unreal Engine 5 is required
Unreal Engine 5 can be downloaded for free via the Epic Games Store; the download is nearly 14GB. The 108GB Zorah TechDemo is available as a 7Z file on GitHub. If you download the engine and demo, however, an RTX 5000 with a Blackwell GPU is also required to run the demo, which utilizes the entire RTX feature set via raytracing.
The hardware shown by Nvidia at GDC in the video was rendered on a GeForce RTX 5090. How does it perform on other Blackwell GPUs? Community feedback is welcome and will be shared over the Easter weekend!
What are neural shaders?
Neural shaders are AI programs within normal shader code. Until now, it wasn’t possible to integrate a neural network into a game without using an API. Neural shaders change that. The AI actions of shader programs are calculated on tensor kernels; the actual rendering is still performed by the normal FP32 ALUS. Microsoft calls this feature “Cooperative Vector.” With neural shaders, in the future, or as in TechDemo, much more realistic worlds can be ruined, as the GPU and graphics memory are relieved by using AI instead of brute-force rendering or pre-made textures. Nvidia divides the neural shader features into three categories:
RTX Neural Materials, RTX Neural Texture Compression, RTX Neural Faces
NVIDIA GeForce RTX 5000 Report: More Details on Architecture, DLSS 3, and Founders Editions.
Another new technology that NVIDIA developers are delivering with the NVIDIA RTX Kit is the new Mega Geometry. The Zorah demo also uses this technology.
Topics: Gaming, GeForce Graphics Cards, NVIDIA Raytracing, Techdemo, Unreal Engine 5
An engineer by training, Alexandre shares his knowledge on GPU performance for gaming and creation.
