A few years ago, Nvidia and AMD, which recently joined Intel, systematically introduced the world of video game algorithms for image optimization and in-game performance with incomprehensible names. Between DLSS, FSR, and XeSS, it’s time to take stock of the solutions available to get the most out of your gaming PC.
In 2018, to accompany the release of the first series of RTX graphics cards, Nvidia introduced the first version of DLSS, which aims to provide a function that improves in-game performance and image quality thanks to various algorithms.
Since then, several new versions have been released (DLSS 3.0 was recently announced by Nvidia) and the competition is now offering its own solutions: FSR from AMD and XeSS from Intel. The principle and promises are more or less the same every time, with some characteristics worth checking out, especially in the midst of bringing new products to market.
DLSS, Nvidia’s exclusive solution
DLSS (Deep Learning Super Sampling) is the first of these solutions introduced by Nvidia in 2018, shortly after the release of the GeForce RTX 2080 and 2080 Ti.
In its first version, its principle is very simple: when you activate it, your graphics card draws game images with a resolution lower than those that will be displayed on the screen. For example, for 4K games, your computer can lower the compute definition to 1080p, which is easier to manage, and then upscale the image to match the screen.
This technique is called Upgradeis useful for achieving performance gains in a very simple way, but has the disadvantage of causing visual effects and creating a less “pretty” and less “sharp” overall image.
To correct this, Nvidia has trained artificial intelligence on its own hardware to create image optimization algorithms. In concrete terms, Nvidia supplies its AI systems with very high definition images, so that they can use it to recreate the missing details. The result of this training is embedded in the models that are then provided to the graphics cards (via their drivers) and implemented by special modules (called Tensor Cores). Because of this, DLSS is only available on Nvidia graphics cards, more specifically on RTX family cards (2000, 3000 and 4000 series).
Finally, the recently announced third version of DLSS is even more accomplished: where previous versions only rebuilt parts of an image sequentially, this version is able to reconstruct an entire image with each processing, allowing according to Nvidia to enjoy additional performance boosts without quality loss. DLSS 3.0 is only compatible with RTX 4000 series and above. In addition, a very limited number of compatible games are currently available.
FSR, AMD Free Response
In the face of Nvidia’s proprietary solution, AMD introduced its own system in 2021, called FSR (FidelityFX Super Resolution). The basic principle is the same as with DLSS: it is again a matter of calculating the game image at a lower resolution and then enlarging it to the screen size, then improving the quality.image thanks to the different algorithms for that end.
Since AMD graphics cards don’t have compute units dedicated to AI models, FSR uses more traditional technologies to process them. This gives it the advantage of working on more card models, including the competition (cards from Nvidia and Intel also benefit from it), but also on older references (some cards released in 2016 are supported).
Recently, AMD also announced a new version of FSR, which is moving to 3.0, and thus will directly compete with DLSS 3.0 for similar use, as this new generation will also use AI-based models for image processing. FSR 3.0 won’t be available until 2023, so not all details are known at the time of writing. However, AMD has already hinted that if this new system will be more efficient on the latest generation of graphics cards (the Radeon RX 7000, which will integrate dedicated AI modules), then FSR 3.0 should still run on the brands’ GPUs. The competition.
XeSS, Intel is entering the jig
Recently, Intel also introduced XeSS (Xe Super Sampling), its own in-game image processing solution, similar to those offered by Nvidia and AMD. It accompanies the timid arrival of its first commercial models of graphics cards (called the Arc Alchemist, but they’re still hard to find on the market).
As with its competitors, the question here is also to calculate the game at a lower resolution in order to increase performance, while compensating for the loss of quality thanks to the image optimization algorithms. Again, the algorithms Intel uses are based on models derived from AI computations, but here they apply using classic GPU compute units, which means XeSS compatibility goes beyond Intel processors and also runs on models from AMD and Nvidia.
Still quite new, this technology looks promising, with performance gains similar to those allowed by FSR or DLSS.
Choose the most appropriate solution
After reviewing the proposed solutions, how do you choose the one that best suits you? First of all, it will depend on two factors that you have to take into consideration.
Your equipment will, at first, adapt the technologies you will have access to. For example, DLSS will only work on Nvidia’s GeForce RTXs, while competing solutions support all GPU brands, though not without limitations.
Typically, FSR and XeSS are generally compatible with GeForces released from 2016 (GeForce GTX 1000 and later), Radeons from 2019 and later (FSR may work under certain conditions on older Radeons), and all Intel Arc GPUs. The game you play will also matter, as developers need to specifically incorporate these solutions into their titles for you to enjoy. Some will offer none, others will offer all three and let you choose. Each manufacturer maintains a list of compatible games on their site.
Note that it is not necessary (and usually impossible) to enable more than one of these solutions at a time. They will all have a performance-enhancing effect, but the result won’t be doubled if FSR and XeSS are activated at the same time, for example.
Thus, the difference between these three methods mainly lies in the quality of the image that you will get in the game. Thus, if you have a compatible Nvidia graphics card, DLSS will always be the best solution if it is offered, and this is especially the case. Optimized for brand specific GPUs. The other two technologies have in common that they are quite the same and are available on all graphics cards currently on the market, and furthermore make it possible to achieve somewhat similar performance gains.
Finally, if your hardware is up to date and your games are compatible, the best solution is still to test to determine which options work best for you. The main goal of these technologies is to improve your device’s performance in the game (to take full advantage of really big advances like ray tracing). But ultimately it all comes down to convenience and personal preference. So feel free to play around with the settings to find the one you like the most, those few wasted minutes will be greatly compensated by your hours of gameplay.