The fourth iteration of Nvidia's Deep Learning Super Sampling technology has arrived. For the first time, it is said to facilitate a frame rate that is more than eight times higher and a latency that is roughly half that of native rendering. As a comparison video from Nvidia (embedded below) shows, the frame rate achieved by DLSS is expected to double once again in the latest version compared to DLSS 3.
Are you a techie who knows how to write? Then join our Team! Wanted:
- News Writer (Romania based)
Details here
To make this possible, Nvidia has integrated "Multi Frame Generation" into DLSS Frame Generation. As the name suggests, DLSS 4 now no longer "only" generates every second frame via AI, but rather three quarters of all displayed frames. Nvidia states that this was not previously possible because it would have put too much strain on the GPU and thus reduced the output frame rate, but thanks to hardware innovations in the GeForce RTX 5000, Multi Frame Generation is now a reality.
The brand new Frame Generation AI model works 40% faster and uses 30% less VRAM. Using Warhammer 40K: Darktide as an example, Nvidia claims that DLSS 4 enables a 10% higher frame rate, even with conventional Frame Generation, while sparing 400 MB of VRAM. In conjunction with GeForce RTX 5000, DLSS 4 is reportedly able to calculate almost 94% of all pixels using AI instead of rendering them natively.
The subsequent effect on performance is clear: As the benchmarks embedded below show, depending on the game, DLSS 4 increases the frame rate by between just under 370% and up to 720%, i.e., by almost fourfold to more than eightfold.
While Multi Frame Generation is only available with the GeForce RTX 5000, users of older graphics cards should also benefit from DLSS 4 (even back to GeForce RTX 2000) since the AI models for regular Frame Generation, DLSS Ray Reconstruction, DLSS Super Resolution and DLSS Edge Smoothing (DLAA) have also been improved.