NVIDIA has announced its biggest efficiency improvement to DLSS 5 since the technology's launch. The October update to its 3D-Guided Neural Rendering model is twice as fast as the September version, according to NVIDIA, and uses 20% less VRAM at 4K. That is meaningful news for a feature whose computational demands have been a major concern.
But speed alone does not resolve the debate surrounding DLSS 5. The technology does more than traditional upscaling: it applies neural rendering to lighting and material appearance. Players and artists want to know whether the result looks better without compromising a game's intended visual identity.
What NVIDIA actually changed
NVIDIA says pipeline optimizations and model refinements reduced the model's footprint by 24%, to approximately 118 million parameters. Its claimed GPU overhead is now half that of the September version while preserving comparable visual quality. NVIDIA also says the model is ten times faster than when DLSS 5 was first announced in March.
Those figures describe the model and its overhead. They do not mean every game will suddenly double its frame rate. Game performance still depends on the GPU, the base render, resolution and how the developer implements the effect.
When can you actually use the faster model?
The first scheduled rollout is NBA 2K27 on 16 October, with NVIDIA claiming frame rates up to 1.4 times higher using the updated model. NVIDIA says the same model will be available on RTX Spark laptops, and it is scheduled for Hogwarts Legacy on 5 November. Where Winds Meet is also on the roadmap. NVIDIA's RTX Remix toolkit is due to receive DLSS 5 on 23 October.
These are vendor announcements and future release dates, not independent performance measurements. Players should check each game's update notes before expecting the new model to appear.
The art-direction question is not going away
DLSS 5 is described as 3D-Guided Neural Rendering, using the engine's rendered image and scene information as a foundation while adding AI-generated lighting and material detail. NVIDIA says developers can tune the result and preserve artistic intent. That control matters, because photorealistic skin or brighter reflections are not automatically more faithful to a stylized game.
Early DLSS 5 comparisons drew criticism when characters appeared visibly altered. NVIDIA's latest Hogwarts Legacy demonstrations reportedly show a more restrained implementation. Whether players prefer it remains a subjective judgment best made from side-by-side gameplay, not marketing stills.
Do not confuse DLSS 5 with the new Frame Generation beta
Alongside the DLSS 5 model update, NVIDIA introduced a separate beta Frame Generation Performance Mode called Preset C. NVIDIA says it uses a lighter optical-flow approach, offering up to 35% faster frame generation in its GeForce RTX 5060 example, with some image-quality trade-offs.
That beta is relevant to supported RTX 40 and RTX 50 Series cards, but it is not the same as enabling DLSS 5 neural rendering. The model, GPU compatibility and settings differ. In particular, NVIDIA's published DLSS 5 FAQ says official RTX 40 Series neural-rendering support remains a future goal, not an already delivered feature.
STATIC's take: better performance deserves better comparisons
The efficiency improvement is substantial if it holds up in independent testing. But the right review should compare actual frame times, input responsiveness, VRAM use, scene stability and artistic fidelity at matched settings. More photorealistic does not always mean more beautiful, and a faster AI model does not automatically make every game more enjoyable.
The most encouraging development may be that developers are refining the effect rather than treating it as a one-size-fits-all filter. The next round of side-by-side gameplay tests will be more useful than any single headline percentage.