Control Rests with Developers in DLSS 5, AMD Keeps the Pressure On

The control side also looks simpler. The main tools explained for developers are centered around two core settings: structure intensity and tone intensity. Structure intensity affects finer details such as ambient occlusion, reflections, and subsurface scattering. Tone intensity works on overall lighting, color, and the scene’s general appearance.

This setup gives developers the ability to produce different results within the same scene. A stronger touch may be preferred in a cinematic, while a more balanced setting can be used in gameplay. In NVIDIA’s demonstration, this model selection can change by scene, environment, and cut. That makes it possible to build a more controlled pipeline instead of applying one filter everywhere.

What does object-based control bring to players?

One of DLSS 5’s most notable features is object-based control. NVIDIA showed that it can mask characters and specific objects separately. If the developer wants, the effect can be enabled on a character while leaving the surroundings unchanged, or different strength levels can be defined for bottles, fruit, or other scene elements.

This level of detail matters for keeping a game’s visual language consistent. Not every scene needs the same processing load. On some characters the effect can be kept lighter, while on certain objects it can be used more prominently. That provides flexibility both for performance and visual consistency.

NVIDIA also makes it clear that DLSS 5 is not designed to replace raster, ray-traced, or path-traced structures, but to expand them. That distinction matters, because the technology does not take the place of the existing render pipeline; it adds to it. For players, that means more detailed surfaces, cleaner reflections, and smoother lighting transitions. For developers, it offers control over how far the image can be processed.

Where does AMD’s Helios move take the competition?

As technical details around DLSS 5 were being revealed, AMD sent another message from its own side. The company announced a new rack system called Helios for large AI labs and data centers. The system is positioned as a direct challenge to NVIDIA’s market dominance.

AMD CEO Lisa Su said Helios is the highest-performance AI system in the technology sector. It is designed to train and run the most complex AI models. NVIDIA’s Vera Rubin and Grace Blackwell systems have held a strong position in this area so far; however, in performance comparisons shared for Helios, it is said to outperform Vera Rubin in some areas.

This race is no longer being fought on raw hardware power alone. AMD’s ties with OpenAI, Meta, Oracle, and Anthropic stand out. Microsoft’s announcement that it will scale its Azure infrastructure with Helios also supports that picture. Anthropic’s statement that it has formed a strategic partnership with AMD for up to two gigawatts of GPU capacity shows just how large the scale is.

Another shift is happening in the background of the technology war. On AMD’s side, conversations about CUDA appear to have largely lost importance. According to the company, discussions with customers on this topic have dropped sharply. That is being read as a sign that NVIDIA no longer holds quite the same unquestioned advantage at the software layer.

What could the short-term result of this race be?

For now, the picture is split into two separate fronts. With DLSS 5, NVIDIA is trying to build a model for image generation that is more controlled, more customizable, and more open to developers. AMD, meanwhile, is putting forward a larger and more forceful scale in AI infrastructure with Helios. One targets the final layer of game visuals, the other targets the data center front. Both are redefining where AI sits on the hardware side.