What is DLSS? Nvidia DLSS, or Deep Learning Super Sampling, is a suite of graphics technologies that looks to improve the frame rate in your games while maintaining high image quality, or even improving image quality through the application of various AI-enhanced technologies.
There have been four main revisions of the technology so far; the original DLSS, DLSS 2.0, DLSS 3, and the latest DLSS 3.5. We’ll explain all four versions below, as well as provide an in-depth explanation of the technologies contained in each. DLSS is rivaled by AMD FidelityFX Super Resolution (FSR) technology.
DLSS started out as a single technology for upscaling games from a lower resolution to a higher resolution, in order to improve frame rates while maintaining decent image quality. However, since the first version of the software it has expanded to incorporate five main features: DLSS Super Resolution, DLSS Deep Learning Anti-Aliasing, DLSS Frame Generation, Nvidia Reflex, and DLSS Ray Reconstruction. We explain each of these further down the page, but first here’s a brief history of each version of DLSS.
The original version of DLSS introduced the concept of DLSS Super Resolution upscaling. However, it was an early version of the technology that required its AI-derived upscaling algorithm to be trained on each game individually by the developers, with thousands of in-game frames compared by the system to create a unique ‘best practice’ for each game. This made DLSS 1 particularly slow and cumbersome for developers to implement and the end result was often not very good, resulting in a blurry image.
DLSS 2.0 introduced an enhanced version of DLSS Super Resolution that uses a generic upscaling algorithm already trained on hundreds of games. This made it easier for developers to implement and resulted in far better image quality.
Where DLSS gets a lot more confusing is the introduction of DLSS 3. It includes the advanced Super Resolution tech of DLSS 2.0, but also introduces a new Frame Generation technology that can work completely independently of Super Resolution. This means games can include AI Frame Generation without DLSS 2.0 Super Resolution. Also, because Frame Generation only works on RTX 40 series cards, such as the GeForce RTX 4070, a game can offer the full range of DLSS 3 features but most RTX gamers won’t be able to use them. DLSS 3 also requires Nvidia Reflex, meaning this existing technology is now partially considered part of the DLSS suite.
DLSS 3.5 includes all the features of previous versions of DLSS and introduces a new Ray Reconstruction technology that improves the image quality of ray traced lighting and shadows for minimal performance hit. It is available to all GeForce RTX GPUs.
Will DLSS work on my graphics card?
All versions of DLSS are exclusive to Nvidia GeForce RTX graphics cards. So they don’t work on any AMD or Intel graphics cards and only work on RTX 2000 series cards, or later, such as the RTX 2060, RTX 3060, and RTX 4060.
The Frame Generation technology introduced in DLSS 3 is exclusive to RTX 4000 series cards. However, somewhat confusingly, the Ray Reconstruction tech of DLSS 3.5 is available on all RTX cards.
Will my game run DLSS?
All versions and features of DLSS have to be incorporated into games by the developers – they’re not options users can switch on in a graphics driver. Also, not all features of DLSS will be used in all games. Nvidia maintains a list of games with DLSS features here.
What is DLSS Super Resolution?
DLSS Super Resolution is the original DLSS tech, which aims to improve frame rates by upscaling games from a lower to a higher resolution.
This allows a game to render the 3D portion of the scene (everything but the menus and HUD) at a lower resolution and then upscale the 3D rendered image to the full resolution of your display. This allows for a faster frame rate while still producing decent image quality.
The main drawback of DLSS Super Resolution is that it can be affected by ghosting when fast movements are made in-game, due to the way it compares previous frames to the current one to inform its upscaling algorithm. That’s why we don’t recommend using it in fast-paced games such as competitive first person shooters. Like all DLSS technologies, Super Resolution is only available on Nvidia GeForce RTX graphics cards.
What is DLSS Deep Learning Anti-aliasing?
DLSS Deep Learning Anti-aliasing (DLAA) reduces the jagged or flickering edges of objects in 3D games.
It does this by using a general technique known as temporal anti-aliasing, which is where the game compares the just-rendered frame with previous ones to help it judge the color of each pixel. DLSS Super Resolution does the same thing – it has built-in temporal anti-aliasing – but with Deep Learning Anti-aliasing there’s no upscaling element.
Like DLSS Super Resolution and many other temporal (time-based or previous-frame-comparison-based) graphics technologies, the main drawback to DLAA is it can produce image ghosting or smearing. It’s again why we don’t recommend using it in fast-paced games. DLAA is available on all Nvidia GeForce RTX graphics cards.
What is DLSS 3 Frame Generation?
DLSS Frame Generation can improve game frame rates by inserting new AI-generated frames in between each normal frame.
Introduced in DLSS 3 alongside the launch of the RTX 4090, it allows for huge increases in frame rate for almost no impact on image quality – though the extra frames are not ‘real’ but made up ones. Like DLAA, it can be used without Super Resolution scaling, so the game renders at full resolution, with the higher frame rate only coming from the extra generated frames.
Again, because the technology has a temporal element it can suffer from visual glitches if there’s fast movement onscreen. It also can affect game latency, which is why DLSS 3 Frame Generation also requires the use of Nvidia Reflex. Also, Frame Generation can only be used on RTX 4000/RTX 40 series cards, making use of a new Optical Flow Accelerator in these GPUs.
What is Nvidia Reflex?
Nvidia Reflex is a technology for reducing overall system latency to ensure games feel as responsive as possible.
Originally a technology separate from the DLSS suite, Nvidia Reflex is now required when using DLSS 3 Frame Generation. Our experience of Nvidia Reflex is that the real-world difference in latency is often imperceptible. However, we’ve also encountered no meaningful downside to turning the option on, if your game has it. Reflex also works on much older non-RTX GPUs, with support going back to at least the GTX 900 series.
What is DLSS 3.5 Ray Reconstruction?
DLSS Ray Reconstruction can be used to enhance the quality of ray-traced lighting and reflections in games.
It does this by comparing previous ray-tracing lighting samples to the current frame’s samples and using some AI-enhanced algorithms to maintain as much detail as possible. The result can be vastly improved ray-traced lighting effects with minimal performance impact.
We’ve yet to test Ray Reconstruction, so we can’t comment on its pros or cons but because it uses a temporal element, it could suffer visual glitches with fast movement. Ray Reconstruction is available on all Nvidia GeForce RTX graphics cards.
Do you regularly use DLSS technologies or do you think they’re all ‘cheating’ and creating ‘pretend’ frames? Let us know your thoughts on the Custom PC Facebook page, via Twitter, or join our Custom PC and Gaming Setup Facebook group and tap into the knowledge of our 400,000+ members.