Crimson Desert’s Lighting Transformed by Nvidia & AMD Tech
▼ Summary
– Crimson Desert’s visual quality is most dramatically improved on PC by using AMD’s ray regeneration or Nvidia’s ray reconstruction, which are far more impactful than standard ray tracing.
– The game uses intensive but heavily optimized ray tracing, running at very low ray counts to maintain performance, which inherently compromises visual fidelity.
– These machine learning-based denoisers transform lighting and reflections, adding accurate shadows, directional light, and stability to moving surfaces that the standard denoiser lacks.
– Adopting this high-quality denoising comes with a significant performance cost, reducing frame rates by 14-24 percent depending on the GPU and settings.
– Both AMD and Nvidia’s solutions have current issues, such as a sub-native image look or graphical bugs, and the pre-launch build also exhibits problems like shadow flickering and pop-in.
The visual impact of Nvidia’s ray reconstruction and AMD’s ray regeneration in Crimson Desert is nothing short of dramatic, offering PC players what feels like a fundamental upgrade to the game’s lighting engine. While the title already uses ray tracing extensively for its global illumination, the standard denoising solution creates a noticeable compromise in quality to maintain performance. Switching to either of these machine learning-powered alternatives radically transforms the scene, enhancing shadows, reflections, and overall lighting fidelity to a degree that often resembles toggling ray tracing itself on or off.
Crimson Desert’s performance across various platforms is achieved through significant optimization. Its surfel-based ray-traced global illumination operates with a very low ray count per pixel, and reflections run at a reduced resolution. A lightweight denoiser cleans up the resulting noise, but this comes at a visual cost. With the standard solution, lighting can appear flat and lack direction, with missing contact shadows and grass that seems poorly lit. The advanced ML denoisers from AMD and Nvidia directly address these shortcomings.
The improvement is immediately visible. Proper, tight shadows appear under geometry like pipes and overhangs, giving objects a grounded, realistic presence within the environment. Directional lighting is restored, and localized light sources gain intensity and clarity. The benefits extend beyond static scenes, as these technologies also resolve issues with dynamic elements. Low ray-count reflections on moving surfaces, such as water, can exhibit a stippled, ghostly appearance with the standard denoiser, making them seem laggy. ML denoising produces a stable, responsive image free from these artifacts. The cumulative effect is so profound it essentially unlocks an ultra-quality lighting tier exclusive to supported PC hardware.
Adopting this higher-quality denoising does carry a computational penalty, resulting in a frame-rate drop. Performance impacts vary by hardware; for example, enabling Nvidia’s ray reconstruction on a high-end GPU at 4K can reduce performance by around 14 percent. The hit on comparable AMD hardware with its ray regeneration enabled can be even more pronounced, sometimes exceeding 20 percent. As with any demanding graphical setting, players will need to balance their preference for visual quality against their desired performance level.
It is important to note that both solutions, while transformative, are not without their quirks in the pre-release build. AMD’s ray regeneration can sometimes struggle to integrate seamlessly with upscaling, leading to a slightly soft or sub-native image quality in certain situations. Nvidia’s ray reconstruction generally delivers a cleaner output but exhibited bugs where surface details from displacement maps were less pronounced, and rain effects could intermittently vanish. The developer is aware of these issues, and improvements are expected.
Other visual quirks exist independently of the denoiser choice, including occasional flickering in sun shadows and noticeable pop-in for some environmental details, even on the highest settings. These may be addressed in the final release code. Ultimately, this showcase underscores how critical advanced denoising has become for ray-traced graphics. The leap in quality provided by these ML techniques is so significant that it sets a new benchmark, highlighting the indispensable role this technology will play in shaping the visual future of gaming.
(Source: Digital Foundry)
