In line with today's release of the VR title Half-Life: Alyx, NVIDIA has released a new game-ready driver for the GeForce graphics cards. He also brings other goodies for other games.
The new game-ready driver for GeForce graphics cards and notebooks brings a bunch of optimizations with it to get Half-Life: Alyx up and running as smoothly as possible and to optimize performance and latencies especially with regard to VR gaming of the title. On the side, players from Resident Evil 3 and Tom Clancy's Ghost Recon: Breakpoint can look forward to various optimizations.
The driver also includes Deep Learning Super Sampling (DLSS) version 2.0 for the games MechWarrior 5 and the RTX / DLSS flagship title Control. In MechWarrior 5: Mercenaries, DLSS 2.0 improves the performance increase of 30 to 75 percent through the quality mode, whereby the image quality is maintained and in some cases even improved.
Regarding control, the upcoming DLC The Foundation, which is due to be released on March 26, will also integrate DLSS 2.0 as part of a patch, so the drivers are already up to date by then. With DLSS 2.0, the image quality is compared to the original DLSS Image processing algorithm used in Control has been significantly improved. Depending on the specific GPU, the resolution and the quality settings, the frame rates in Control can be increased by up to 75 percent.
DLSS 2.0 includes general process optimization. It offers image quality comparable to the native resolution, although it only renders a quarter to half of the pixels. The new AI network uses Tensor Cores even more efficiently to perform calculations 2x faster than the original. This improves the frame rates and eliminates previous restrictions regarding settings and resolutions on certain GPUs.
The original DLSS required training the AI network for every new game. DLSS 2.0 does not train with the content of specific games and thus provides a general network that works across games.
In this clip, NVIDIA explains the features of the new version of DLSS (Deep Learning Super Sampling).
