For the first time in years, Nvidia opted not to launch new GeForce graphics card models at CES. CEO Jensen Huang’s extensive and seemingly impromptu 90-minute keynote was predominantly oriented towards the company’s leading AI business, pushing gaming-related announcements to a separate video release later in the evening.
In lieu of new hardware, Nvidia emphasized software enhancements for existing devices. The most notable update is DLSS 4.5, which introduces a suite of new features to Nvidia’s array of upscaling and frame generation technologies.
The DLSS upscaling has been bolstered by a novel “second-generation transformer model” that Nvidia claims has been “trained on an expanded data set” to enhance pixel generation predictions. Bryan Catanzaro from Nvidia highlighted that this is especially advantageous for image quality in the Performance and Ultra Performance modes, where the upscaler faces challenges due to working with lower-resolution source images.
Moreover, DLSS Multi-Frame Generation has been upgraded, increasing the AI-generated frames per rendered frame from three to five. This innovation includes a new 6x mode for DLSS MFG, complemented by a feature called Dynamic Multi-Frame Generation, which allows the number of AI-generated frames to vary. This adaptation increases generated frames during intense scenes and decreases them during simpler ones, ensuring computations are efficient.
The usual limitations for Multi-Frame Generation apply: it necessitates an RTX 50-series GPU (as the 40-series GPUs can only generate one frame per rendered frame and older cards cannot generate additional frames), and the game needs to run at a sufficiently high base frame rate to avoid lag and peculiar artifacts. While useful for enhancing performance in fast-spaced games, it remains ineffective for rendering unplayable frame rates into playable ones.