NVIDIA Introduces DLSS 3.0 Exclusive to RTX 4000 GPUs, Offers A 4x Increase In Performance

By Abdullah FaisalSeptember 20, 2022

2 minutes read

Recently, NVIDIA has announced their new RTX 4000 series based on the ‘Ada Lovelace‘ architecture. In addition to that, a new DLSS 3.0 has also been announced offering up to a 4x increase in performance. However, that comes at the cost of you buying a new Lovelace GPU.

How DLSS Works?

To put it simply, DLSS renders your game at a lower resolution. With the power of AI fueled by NVIDIA’s Tensor cores, that image is upscaled at a desired resolution. The result is you get a close-to or sometimes even better image than native res.

What DLSS 3.0 Bring To the Table

NVIDIA’s DLSS 3.0 is a major uplift as compared to DLSS 2.0. The former uses a new AI powered Optical Multi Frame Generation system rendering entire new frames, instead of pixels (DLSS 2.0). DLSS 3.0 leverages a combination of the new ‘Super Resolution’, ‘DLSS Frame Generation’ and ‘NVIDIA Reflex’ to reconstruct 7/8th of the displayed pixels.

DLSS is powered by NVIDIA’s 4th Gen Tensor cores and the Optical Flow Accelerator shipping with Ada Lovelace making it RTX 4000 exclusive.

How DLSS 3.0 Works

The Optical Flow Accelerator in Ada Lovelace tracks each and every pixel down to the last bit. Even details such as reflections, shadows which are not shown by the game engine are also traced to calculate the ‘motion flow’. This allows DLSS 3.0 to reconstruct an entire frame pixel by pixel.

If DLSS 3.0 used the in-game motion vector engine which does not calculate the motion of shadows relative to the object’s position. Then shadow stuttering would occur as shown in the example below:

DLSS 3.0 is a combination of both the game’s vector motion engine and NVIDIA’s Optical Flow. The DLSS Frame Generation AI network calculates how much information to use from both sources, thus constructing the perfect image.

The first frame is rendered at a downscaled resolution with NVIDIA’s super resolution increasing the image’s quality to match native-res. The second frame is completely constructed by NVIDIA’s Frame generation network. In short, NVIDIA states that almost 7/8th of the image is generated by AI, drastically reducing the burden on the GPU.