The latest in Nvidia’s DLSS family of technologies, the DLSS 3, was released on October 12 along with the GeForce RTX 4090 GPU. While the previous two generations are available on all of their RTX graphics cards, so far it looks like DLSS 3 might only be available for their RTX 40- series. This series, named after English mathematician Ada Lovelace, was earlier announced on September 20 and is currently used in the GeForce RTX 4090 and 4080 GPUs. This release from one of the GPU giants of the world has brought much excitement, as they promised frame rate improvements of up to 4x.
In a nutshell, DLSS 3 is reportedly even more powerful than its predecessors, allowing players to play games at a high resolution without compromising on frame rate or visual details. This proves to be less of a strain on the graphics card, but the process by which this happens is a bit more complex. Let’s take a look at how this technology works.
What is DLSS?
DLSS stands for Deep Learning Super Sampling and is a feature of Nvidia’s modern GPUs. Essentially, a low-resolution image is generated, which is then informed by the ‘deep learning’ part of the technology. Deep learning refers to the use of AI to fill in missing visual information in a low-resolution image so that the game performs as if it was rendered at this higher resolution. The AI is programmed with high-resolution scans, which are then used by anti-aliasing methods to improve the rendered image. This avoids the drawbacks of locally rendering high-resolution images, reducing the strain on your computer’s hardware.
- Chipset: GeForce RTX 4090
- Video Memory: 24GB GDDR6X
- Memory Interface: 384-bit
- Output: DisplayPort x 3 (v1.4a) / HDMI 2.1 x 1
- Digital maximum resolution: 7680 x 4320
How does DLSS work?
By drawing information from AI trained with super-resolution images, DLSS can generate a lower-resolution image so the game feels like it’s running at a higher resolution than it actually is. Although the AI is trained with extremely high-resolution images from certain games, this information is compressed into mere megabytes before being loaded onto Nvidia’s drivers. This technology is designed to work with and learn from visual artifacts and bugs present to sharpen the final image even further, as opposed to traditional technologies. DLSS 2, now known as DLSS Super Resolution, offers four times the resolution of DLSS, making it possible for games rendered at 1080p to seem like they’re running at 4K.
What makes DLSS 3 different?
The newest iteration of this technology brings some significant changes, mainly due to the addition of optical flow analysis. According to Nvidia, game performance is enhanced by up to 4x compared to traditional rendering, and the RTX 40- GPUs render up to twice the frame rate the CPU can compute. The algorithm works by taking two frames rendered by the GPU and generating a new frame that transitions between them seamlessly. This is in contrast to DLSS and DLSS 2, which only use AI to reconstruct sharper images from those rendered with fewer pixels, as opposed to creating new frames entirely.
Usually, DLSS 3 doesn’t have worse visual quality than DLSS 2, but unfortunately, this isn’t always the case. Certain user-interface elements can become jumbled due to the use of AI-generated frames, leading to distorted images and sometimes unreadable text, such as in the case of Cyberpunk 2077. The frame rate was still substantially increased, however, just like was promised – Nvidia’s DLSS 3 provided almost 50% more frames compared with DLSS 2.
Worth noting is that Nvidia intends for DLSS 3 to be used alongside their Reflex technology, aiming to mitigate the extra latency added through the generation of these additional frames and reduce lag. The lower latency makes on-screen actions more instantaneous and game controls more responsive. DLSS 3 reduces latency by up to 2x compared to the original DLSS.
- NVIDIA Ada Lovelace Streaming Multiprocessors: Up to 2x performance and power efficiency
- 4th Generation Tensor Cores: Up to 2X AI performance
- 3rd Generation RT Cores: Up to 2X ray tracing performance
- Axial-tech fans scaled up for 23% more airflow
- Dual Ball Fan Bearings last up to twice as long as conventional designs
Will DLSS3 come to older GPUs?
While DLSS 2 is available for over 216 games and apps, DLSS 3 is currently being supported by around 35, with the first of these being launched in October 2022. Although there’s technically no reason why DLSS 3 can’t be run with GPUs older than Ada, the idea has been somewhat shot down by Bryan Catanzaro, vice president of Applied Deep Learning Research at Nvidia. He explained on Twitter that DLSS 3 relies on the optical flow accelerator, which ‘has been significantly improved in Ada over Ampere’, reportedly being 2 to 2.5x faster. This is why players likely won’t see much of a frame rate boost on Ampere or Turing GPUs, and may even see subpar image fidelity and experience lag-filled gameplay.
In theory, Nvidia could improve DLSS 3 to allow it to run faster on older hardware, but there is not much incentive currently for them to do so. It doesn’t look like we will be receiving full compatibility in the near future, but there is always hope.
Last update on 2022-11-30 / Affiliate links / Images from Amazon Product Advertising API