NVIDIA has officially pushed out the latest DLSS update tech, namely the DLSS 4.5 Super Resolution for everyone using the NVIDIA app.
The 2nd-gen transformer model is now applicable for more than 400 games and apps for all GeForce RTX cards; however, this utilizes even more hardware power than usual, 5 times at that, and it exclusively uses FP8 data types, which are not supported on the older RTX 20 and RTX 30 series cards, it could technically result in negative results rather than positives.
However, if you’re running the latest RTX 50 series GPUs, you’ll be able to enjoy cleaner image quality and noticeably better temporal stability, especially when running DLSS Performance and Ultra Performance modes where artifacts usually show up first.
There’s also the 6x Multi Frame Generation for even more FPS under heavy settings to make games look better, with upcoming titles like 007 First Light, Phantom Blade Zero, PRAGMATA, Resident Evil Requiem, Active Matter, Screamer, and DEFECT – all getting ready to deliver next-level graphics and gameplay.
On the other hand, this week sees LET IT DIE: INFERNO getting an update that adds DLSS Super Resolution, with the NVIDIA app letting players override it to DLSS 4.5 for even better results, while StarRupture launched into Early Access with DLSS Super Resolution and Frame Generation right out of the gate, and can be pushed further with DLSS 4.5, DLAA, and Multi Frame Generation.
Since it’s about the NVIDIA App providing access DLSS 4.5 in some way, the program is now updated to version 11.0.6 that brings over the remaining NVIDIA Surround options from the old Control Panel, including bezel correction and hotkeys, laptop users with Advanced Optimus now get clearer visibility on which apps are blocking the feature instead of just seeing vague .exe names, and there’s a new Debug Mode that instantly disables all overclocking and voltage tweaks to help troubleshoot stability issues without digging through multiple apps.
For devs, check out the new DLSS 4.5 Super Resolution Streamline plugin, as well as the RTX Neural Texture Compression SDK version 0.9 that packs BC7 encoding that is 6x faster as well as 40% higher inference speed alongside reduced memory overhead with minimal visual loss.
Other notable additions include expansion of NVIDIA ACE with new small language model options like Nemotron Nano 9B V2 and additional Qwen3 variants, plus a new Nsight Graphics 2025.5 release focused on better debugging and profiling workflows.
