What is NVIDIA DLSS & How Does it Work?

Introduction

Suggested Article: What is a GPU and Do You Need One For Gaming?

What Does DLSS Stand For?

DLSS stands for Deep Learning Super Sampling. The tech uses deep learning AI in order to render games at a lower native resolution and upscale to your displays resolution. This essentially gives you the frame rate levels you would find at a lower resolution, but the visual fidelity of the higher, display resolution.

AI-accelerated super sampling (a type of anti aliasing, leveraged to remove jagged edges in games) is then used to ensure the best visual fidelity possible and reduce the noticeability of the process which has just occurred.

Now, this may seem too good to be true – free frame rate, what’s the catch? With DLSS 1.0, the catch was the quality – jagged edges and a softer image were definitely present. But Nvidia has really pulled through with DLSS 2.0, creating a ground-breaking piece of tech for gamers. The results speak for themselves, and help put Nvidia ahead of AMD with cards that actually boast the same raw power. It’s all about game and software optimisations.

Traditional anti-aliasing methods can cause load to be quite high. A method like Super Sample Anti-Aliasing, or SSAA, renders out an image higher than native, and then scales it down. The issue with doing it this way, is the performance hit is quite severe. Each pixel is essentially scaled up 2x or 4x the native resolution to provide sharper colours and depth.

Is DLSS Proprietary to Nvidia?

Yes. AMD have launched their FidelityFX Super Resolution tech in a bid to rival DLSS – but the results still have some way to go. The main reason why Nvidia’s DLSS provides such a performance boost is because of Nvidia’s secret sauce – ‘deep learning’. Nvidia has trained AI models with machine learning to recognise and process the data of high-resolution scans. By doing this, the anti-aliasing in your game can use the AI model to fill in the data. This provides great super-sampling, but without the overhead to go with it. DLSS also does this offline, minimising the performance loss and increasing frame output.

Alternative Solutions

Currently DLSS provides incredible metrics and opens up higher resolution opportunities to lower-end cards. There are alternatives to DLSS, such as the aforementioned AMD FidelityFX Super Resolution is great competition versus Nvidia’s solution. Games tend to roughly perform at the same level, and this option is also open to Nvidia’s 10 series cards. Intel is also developing an alternative form of supersampling known as XeSS which is an upcoming feature for Intel Arc Alchemist GPUs. In light of new graphics cards being released quickly, manufacturers providing software solutions to increase performance is great to see.

What Cards and Games Can use it?

DLSS is incredible for higher resolutions, but unfortunately there is a caveat. For those of you still using GTX 10 series cards, DLSS is not available. DLSS uses the Tensor cores in the RTX 2000 and 3000 series cards, making these cards a firm requirement to use the tech. DLSS will perform better on the 30 series cards because the new Tensor cores offer better per-core performance, but RTX 2000 series still perform very well. After all, who is saying no to free frame rate?! It is a shame to see DLSS not available on older generation cards, especially as the Steam Hardware Survey shows GTX 1060 cards as the most popular in the market, but the lack of Tensor Cores on 10-series makes this unviable.

In terms of game titles, DLSS is widely supported. Since Nvidia’s release of DLSS the list has only been growing, with most of the popular titles supporting the tech, but there are some gaps. Nvidia has provided a table of their supported games that you can check out, which will give you an idea of what’s around. If your favourite game isn’t on there, don’t lose hope as Nvidia may support in the future!

Performance Metrics

Below we’ve included three games that support DLSS and the kind of frames that you should expect with each game using a mid-range card- the 3060Ti. The resolution is set to 1440p resolution and high graphical settings to really show the power of DLSS.

Battlefield 2042

Starting off, with DLSS on we’re off to a great start! In Battlefield, the 3060Ti achieved an average of 90-100FPS. Battlefield is a rather large game, and in terms of graphical fidelity, it really pushes performance. In areas with large objects and particle effects DLSS really strives in outputting high frames. Without DLSS we’re averaging around 80-90FPS. Although the performance increase isn’t necessarily huge, under areas that cause high stress the fps difference will be noticeable.

Call of Duty Vanguard

In Vanguard the performance increase is rather substantial. Frames with DLSS tend to average around the 170-180 mark, sometimes even 190 upwards. Without DLSS we were only getting around 100 frames on average. We also saw a decrease in colour depth and sharpness. DLSS increases graphical fidelity with the AI rendering, so its disappointing to see less colour without DLSS.

Fortnite

Generally Fortnite is quite a well optimised game hitting well over 120FPS without the need for DLSS, but Nvidia’s solution soars in this instance. From even within the menus with DLSS enabled Fortnite skyrocketed to 270 frames. When we jumped from the Battlebus, the frames leaped to highs of 370. DLSS has really proved its optimisation of performance in Fortnite and will be a great experience for all players!

Additional Titles

DLSS is supported in many titles, so below we’ve got a graph of some of the performance you should expect out of a 3060 Ti and other graphics cards. To briefly describe, on average we’re able to hit relatively good performance in a lot of the triple A titles. The 3060 Ti hits well over 100fps at 1080p High settings, and will jump up to even higher numbers with DLSS enabled. Overall it is incredible to see the level of performance coming out of a mid-range card, especially with software that boosts that performance even higher!

Conclusion

Nvidia’s DLSS is a spectacular integration that massively improves fidelity and performance. On top of that, it opens up an experience for users that don’t necessarily have the hardware for higher settings or higher resolutions. It has also been fantastic to see Nvidia include the Tensor cores needed for DLSS on their latest budget release, the RTX 3050. Giving DLSS support to budget gamers is a major step forward and a positive sign of things to come!

Video