A sizzling potato: Upscaling has change into a subject of competition since Nvidia and AMD launched DLSS and FSR, with some customers viewing the performance as a crutch for unoptimized video games. After unveiling new GPUs that rely closely on picture reconstruction and body era for his or her claimed efficiency positive factors, Nvidia revealed an information level suggesting that the majority RTX GPU customers activate DLSS, which means to point that upscaling has change into the norm.
Nvidia held a prolonged presentation at CES 2025 outlining the previous and way forward for its DLSS neural rendering know-how. One slide included an information level claiming that over 80 p.c of customers with RTX 20, 30, and 40 sequence graphics playing cards use DLSS on video games, vindicating the corporate’s in depth use of AI in online game rendering.
Whereas the slide doesn’t specify how Nvidia collected the information, if correct, it means that players have overwhelmingly embraced machine learning-based upscaling. The presentation additionally highlighted that over 500 video games and 15 of 2024’s high 20 titles help DLSS. Nevertheless, it didn’t make clear whether or not most RTX 40 sequence homeowners particularly use body era, which is a part of the DLSS suite however is a separate step than upscaling.
Upscaling applied sciences like DLSS, FSR, and XeSS render video games under a show’s native decision and use superior methods to reconstruct the lacking pixels, considerably boosting efficiency. Analyses from TechSpot and different retailers display that the elevated body charges typically far outweigh the minor reductions in picture high quality. Nvidia’s newly revealed DLSS 4 goals to reduce these visible flaws even additional by utilizing GenAI transformer fashions.
Nevertheless, as video games like Remnant II, Alan Wake II, and Monster Hunter Wilds now checklist upscaling as a part of their system necessities, debate persists over whether or not the visible trade-offs are justified.
Moreover, Nvidia is selling body era, which interpolates AI-generated frames between historically rendered ones with out lowering latency, as a key characteristic of its RTX 50 sequence playing cards.
Nvidia’s CES presentation additionally outlined future plans for neural rendering, which is able to combine AI-assisted methods into inside rendering pipelines as an alternative of simply resampling the ultimate rendered body. Within the coming years, DLSS might allow video games to course of extra detailed supplies, hair, facial expressions, and different belongings with minimal efficiency prices.
Nvidia has beforehand argued that as efficiency positive factors from standard rendering and semiconductor die shrinks decelerate, enhancing visible element and body charges will more and more rely on applied sciences like neural rendering and body era. These presently tackle the substantial computational calls for of 4K decision and ray tracing.
Furthermore, AMD’s and Intel’s efforts to play catch-up with Nvidia as an alternative of offering different options additional validate the rising significance of AI rendering. Staff Crimson unveiled FSR 4 upscaling at CES, too, which mimics Nvidia’s machine studying method and achieves noticeably higher outcomes than FSR 3.
Equally, late final yr, Intel launched new GPUs that help body era utilizing XeSS 2. The shift towards AI-assisted graphics will probably require new approaches to technical evaluation and benchmarking.