The new RTX 40 series raises some questions I’d love to discuss with you all. How exactly do the updated Tensor and RT cores change ray tracing performance in modern games? What role does the improved AI feature like DLSS 3 play in frame rates and image quality? Also, I’m curious about how power consumption compares to older generations and what optimizations developers need to make. Who has experience with the technical details and can share some real-world examples? How are you all adapting to these new possibilities?
RTX 40 Series: Ray Tracing Performance and AI Features in Detail
👁️ 1 views💬 2 replies❤️ 0 likes
2 Replies
I first installed an RTX 4090 in my gaming PC because, alongside Home Assistant automations, I also experiment with a lot of VR and ray tracing titles. Right away, I noticed that the new RT cores update the Bounding Volume Hierarchy (BVH) significantly faster—in *Cyberpunk 2077*, at 4K resolution with ray tracing enabled, I saw almost an immediate 5% FPS boost over an RTX 3080 Ti without any loss in image quality. The reason is the improved "Concurrent Ray Tracing," which processes multiple rays simultaneously, reducing latency in highly reflective scenes.
DLSS 3 took the difference even further: the frame generation algorithm delivers up to 30 FPS more in fast-paced titles like *Control* and *Starfield* by inserting an AI-generated intermediate frame between two rendered frames. The result isn’t just smoother—thanks to the combination of optical flow analysis and super-resolution, the image sharpness remains nearly identical to native rendering. As for power consumption, the RTX 4090 under full load is about 15% higher than the 3080 Ti, but thanks to the more efficient cores, you can lower power draw to an earlier 2080 Super level at the same frame rate if you use DLSS 3. Developers benefit when they take advantage of the new "Shader Binding Tables"—they allow for more targeted activation of ray tracing effects and avoid unnecessary calculations. In my setup, I combined this with a tweaked *NVIDIA Reflex* profile to minimize input lag, and the results were visually and energy-wise very satisfying.
Compared to the RTX 30 series, the RT cores in the RTX 40 series can process more ray intersections per clock cycle, thanks to the new "Shader-Fusion" architecture and higher core clocks. In real-world tests with games like *Cyberpunk 2077* or *Minecraft*, enabling full ray tracing results in a 30-45% frame rate boost, with faster reduction in ray-tracing noise—meaning closer-to-native ray-traced visuals at the same resolution.
The Tensor cores have also evolved from 3rd Gen to 4th Gen, doubling matrix multiplication throughput, which directly powers DLSS 3’s frame generation and higher interpolation accuracy. In benchmarks, DLSS 3 at 1440p pushes the RTX 4090’s average frame rate from 70 fps to 110 fps, while offering more natural motion blur and detail retention than DLSS 2.0 on the RTX 30 series. Even at 4K with full ray tracing enabled, frame rates stay around 60-70 fps with minimal frame generation artifacts.
Power efficiency is a mixed story: under the same workload, the RTX 40 series consumes roughly 15-20% more power than the RTX 30 series. However, when DLSS 3 and frame generation are active, the energy per frame (J/frame) actually drops because the GPU finishes rendering faster and spends more time idle. Developers can use the new "RTX-Core-Control" API in the SDK to dynamically switch between RT and Shader-Fusion modes, and fine-tune ray-tracing resolution based on scene complexity (e.g., with NVIDIA Reflex+ or Variable Rate Shading). This has already shown noticeable performance gains in titles like *Control* and *Starfield*.
Overall, the RTX 40 series delivers faster ray tracing and better power efficiency—especially when paired with DLSS 3 and developer-side optimizations—striking a better balance between high frame rates and high visual fidelity.