Experimental OptiScaler Mod Cuts DLSS 5 Neural Rendering Performance Hit by Running It Before Super Resolution
A new experimental OptiScaler fork is giving NVIDIA GPU users a promising way to reduce the heavy performance cost of DLSS 5 Neural Rendering. By changing where Neural Rendering is applied in the rendering pipeline, the mod can reportedly cut the frame-rate penalty nearly in half while keeping much of the visual benefit intact.
DLSS 5 Neural Rendering has generated plenty of excitement because of its ability to improve lighting, shadows, and other visual elements in supported games. The downside is that it can be extremely demanding, even on powerful graphics cards. In some cases, enabling the feature can cause a major drop in FPS, making the visual upgrade harder to justify for players who prioritize smooth gameplay.
The new experimental OptiScaler-DLSSNR-PreSR-Multipass fork attempts to solve that problem by moving DLSS 5 Neural Rendering to an earlier stage of the rendering process. Normally, Neural Rendering is applied after DLSS Super Resolution has already upscaled the frame. That means the Neural Rendering model is working on the final higher-resolution image, which requires significantly more processing power.
With this experimental approach, DLSS 5 Neural Rendering runs before DLSS Super Resolution. In practice, this means the model processes a lower-resolution input frame first, such as a 1920×1080 image when using 4K DLSS Performance mode. After that, DLSS Super Resolution upscales the image to the target resolution.
This change may sound simple, but it can make a major difference. Processing a lower-resolution image requires less GPU workload than applying Neural Rendering to a fully upscaled 4K frame. The concept is similar to the way NVIDIA’s DLSS 5 Autopilot pipeline is designed to improve efficiency.
Early user testing suggests the results can be impressive. One Reddit user reported testing the experimental OptiScaler fork on an RTX 5070 Ti in Resonance: A Plague Tale Legacy. According to the user, the game was running at close to 180 FPS in 4K using DLSS Performance mode without DLSS 5 Neural Rendering enabled.
After turning on DLSS 5 Neural Rendering in the traditional post-upscaling configuration, performance dropped sharply to around 80 FPS. That represents a performance loss of more than 50%, which is a huge trade-off even for improved image quality.
However, after configuring the experimental OptiScaler build to run DLSS 5 Neural Rendering before DLSS Super Resolution, the performance improved significantly. The reported frame rate climbed to around 122 FPS, reducing the performance penalty to roughly 25–30% instead of more than 50%.
That kind of improvement could make DLSS 5 Neural Rendering much more practical for many players, especially those using high-resolution displays or demanding graphics settings. Instead of choosing between better visuals and smooth gameplay, this approach may offer a more balanced middle ground.
Image quality remains the biggest question. The user who tested the mod said the visual difference was not significant enough to outweigh the performance gains. For that player, the smoother frame rate made the pre-Super Resolution method the better option.
Other users, however, noted that there may be visible differences compared to the standard DLSS 5 Neural Rendering implementation. Since the model is working on a lower-resolution image before upscaling, some scenes may show reduced detail or altered visual behavior. For players who are highly sensitive to image quality, that could be a deal breaker.
Still, for many PC gamers, the trade-off may be worth it. A jump from around 80 FPS to more than 120 FPS can dramatically improve responsiveness and overall smoothness, especially in visually demanding games. If the loss in image quality is minor, running DLSS 5 Neural Rendering before Super Resolution could become a preferred option for performance-focused users.
It is important to remember that this OptiScaler fork is unofficial and experimental. Results may vary depending on the game, GPU, resolution, DLSS mode, and graphics settings. Stability, compatibility, and image quality may also differ from one title to another.
Even so, the mod shows how quickly the PC gaming community is finding creative ways to improve modern upscaling and rendering technologies. In just a short time, modders have managed to bring newer DLSS features to more hardware configurations and experiment with optimizations that can make demanding features more usable.
If further testing confirms these early results, the idea of applying DLSS 5 Neural Rendering before DLSS Super Resolution could become an important performance-saving technique. For now, it remains an experimental workaround, but it is already attracting attention from players who want the visual benefits of Neural Rendering without sacrificing too much FPS.
