Tech
Crimson Desert’s Lighting Revolution: Why ML Denoising is the Future of Ray Tracing
Crimson Desert, the upcoming open-world RPG, isn’t just pushing boundaries with its expansive world and action-packed gameplay. It’s also becoming a showcase for the next generation of graphics technology, specifically the transformative power of machine learning (ML) denoising in ray tracing. While ray tracing itself delivers impressive visuals, the real leap forward comes from technologies like AMD’s Ray Regeneration and Nvidia’s Ray Reconstruction.
The Compromises of Ray Tracing and the Rise of Denoising
Achieving performant ray tracing often requires compromises. Crimson Desert utilizes a unique surfel-based ray-traced global illumination (RTGI) that operates at a remarkably low ray count – as little as 1/16 rays per pixel for indirect lighting. Reflections are rendered at quarter resolution. This efficiency comes at a cost to visual fidelity, resulting in potentially flat lighting and a lack of detailed shadows. This is where denoising steps in.
Traditionally, denoising techniques have been computationally lean but limited in their ability to fully restore image quality. However, ML-based denoising is changing the game. These technologies reconstruct cleaner lighting data from fewer ray samples, effectively boosting visual quality without the performance hit of increasing ray counts.
AMD Ray Regeneration vs. Nvidia Ray Reconstruction: A Side-by-Side Look
Both AMD’s FSR Redstone Ray Regeneration and Nvidia’s DLSS Ray Reconstruction dramatically improve the visual experience in Crimson Desert. The difference isn’t subtle; it often feels like toggling ray tracing on and off. With the standard denoiser, lighting can appear flat and geometry lacks proper contact shadows. Grass can even appear unlit.
With ML denoising enabled, shadows appear under overhangs, lighting becomes directional, and objects sense grounded in the environment. Reflections on moving surfaces are stabilized, eliminating the ghosting effect seen with lower ray counts. Localised lighting and reflections are also significantly enhanced. The result is an “ultra-quality lighting setting” exclusive to PCs equipped with these technologies.
However, this visual upgrade isn’t free. Enabling these features does impact performance. On an RTX 5080 at 4K performance mode, ray reconstruction results in a 14 percent frame rate drop. The impact is even more pronounced on AMD hardware, with the RX 9070 XT experiencing a 24 percent hit when paired with FSR 4 upscaling.
Current Limitations and Ongoing Improvements
While transformative, both technologies aren’t without their quirks. AMD’s Ray Regeneration currently doesn’t integrate upscaling with denoising, leading to a noticeably sub-native look in some instances. Nvidia’s Ray Reconstruction, in a pre-launch build, exhibited bugs affecting displacement maps and occasionally causing rain to disappear. Pearl Abyss is aware of these issues and working on improvements.
Other visual issues, such as flickering shadow maps and pop-in, were also observed in the pre-launch build, but these may be addressed before the final release.
The Broader Implications for the Future of Gaming Graphics
Crimson Desert highlights a crucial shift in the landscape of gaming graphics. Denoising, particularly ML-based denoising, is becoming as important – if not more so – than simply increasing ray counts. This has significant implications for hardware manufacturers and game developers alike.
Expect to see increased investment in ML-powered graphics technologies. Future games will likely rely heavily on these techniques to deliver visually stunning experiences without requiring exorbitant hardware specifications. The ability to achieve high-fidelity visuals with fewer resources will be a key differentiator in the years to come.
Frequently Asked Questions
- What is Ray Regeneration?
- Ray Regeneration is an ML-based denoiser developed by AMD as part of the FSR Redstone stack. It improves ray-traced image quality by reconstructing cleaner lighting data from fewer ray samples.
- What is Ray Reconstruction?
- Ray Reconstruction is Nvidia’s ML-based denoiser that achieves a similar goal to Ray Regeneration – enhancing ray-traced visuals with fewer resources.
- Will these technologies perform on all graphics cards?
- Ray Regeneration requires AMD RDNA4 architecture. Ray Reconstruction is available on Nvidia RTX series cards.
- Do these features impact performance?
- Yes, enabling ML denoising does result in a performance hit, typically ranging from 14% to 24% depending on the hardware and settings.
Crimson Desert’s embrace of these technologies signals a pivotal moment in the evolution of gaming graphics. The future isn’t just about more rays; it’s about smarter rays, and the power of machine learning to unlock their full potential.
