NVIDIA DLSS – With great graphics comes great performance

Made in partnership with NVIDIA

by Kit

in All, News, Tech

Reading Time: 5 mins read

If you’re gonna be dubbing yourself a bonafide Gamer with a capital G, naturally you’d be wanting that shiny maximum settings, ray tracing, high frame rate, the whole works. Here’s where NVIDIA comes in, with their Deep Learning Super Sampling (DLSS) technology. With over 55 games and counting, including game engines, supporting DLSS, it begs the question: how does it work?

HOW IT WORKS

The simple way to describe NVIDIA DLSS would be to call it a deep learning AI algorithm, where it takes a lower resolution image, then generates an upscaled image based on what it thinks it should be. For your pictures, think of it as being amazingly detailed, with a really tiny file size in comparison, as a very rough idea. DLSS applies that across the board, where the game you’re running doesn’t have to start at its best quality, but technology helps it reach a higher level. You probably won’t expect any random blade of grass to have the same loving detailing as Norman Reedus’ face in Death Stranding, but DLSS means that both things could be equal in quality.

Towa and the Guardians of the Sacred Tree Preview – Narrative Heavy Roguelite Offers Fresh Take on Genre

[BIC2025] Exclusive Interview with Tripearl Games CEO on Adapting The Player Who Can’t Level Up for Gaming

In processing the inputs, DLSS also takes multiple samples of the same pixel to deliver better detail and edge quality. The output comes from four different inputs:

The base lower resolution image from the game engine
Motion vectors of the image from the game engine, which tell the algorithm the direction objects are moving in from frame to frame
High resolution output of the previous DLSS-enhanced frame
A large data set of 16K resolution ground truth images of different game content used in training the AI network

Follow us on Twitter Follow @GamerBraves and Tweet us Tweet to @GamerBraves

The technicalities sees the algorithm predicting pixels: 4K DLSS output is 8.2 million pixels, while 1080p is 2 million pixels, requiring an additional 6.2 million pixels. Drawing from a database of high resolution images, it learns how to best turn that lower quality frame to a much better looking one. All this is done without taxing your PC, for the best of both worlds: no performance loss while enjoying crisp graphics.

For comparison’s sake, take NVIDIA’s own Image Sharpening solution (NIS). NIS is a more traditional upscaling method that does not utilize AI, therefore not having that database that it may learn from. As such, NIS can only average an existing frame’s pixels and throw on a shader: while the result can certainly look good, it cannot reach the heights that DLSS can offer. As the AI continues to learn and improve, it’ll be able to better process more complicated pixels, like things in motion or fog effects, at a speed where hand-coded AI just can’t match.

DLSS AND YOUR GAMES

Fortuitously enough, since its initial introduction in 2019, NVIDIA DLSS is currently in its 2.0 iteration, doubling its efficiency on the Tensor cores it runs on. It’s also customizable, for more user choice and correspondingly bigger performance boosts, just to pick at a couple of its improvements.

In DOOM Eternal‘s case, your machine won’t be having a meltdown when the gore and sparks quite literally begin to fly when you begin revving up your guns and the demons start swarming to meet their doom. Maybe you just want to preen in the ray-traced surfaces, admiring the Doom Slayer’s physique in Photo Mode. Those are options open to you.

HIGH FRAME RATE, HIGH IMAGE QUALITY

While ray tracing is able to add that increased realism to games, often times, it can take a toll on your machine. Not so with NVIDIA DLSS. DOOM Eternal is one of the latest titles to have DLSS support, so what does this mean in conjunction with ray tracing? You get up to 60% accelerated performance at 4K, with ray tracking enabled and all settings at max. Virtually every GeForce RTX GPU can exceed 60 FPS with DLSS on, so that you can enjoy those glorious enhanced visuals with buttery smooth performance without an immediate hardware upgrade.

See those reflections on the floor? That’s thanks to ray tracing, and ordinarily, the sheer processing power needed makes it a really good way to stress test hardware to the limit. DLSS eases this processing load, so power doesn’t need to be concentrated onto just one area, and keeps your ray-traced games looking stunning without sacrificing performance.

ECONOMICAL HIGH PERFORMANCE

As more and more game engines – Unreal Engine, Unity, and other first party engines – support DLSS, it becomes easier for gamers to get that technological boost for their graphics and frame rate while still keeping the required processing power on the lower end, from AAA games to any indies that opt in. That bears repeating, as DLSS technology can make high quality performance that much more accessible even without a top-of-the-line card, though granted the current iteration is only for the GeForce RTX 20 and RTX 30 series GPUs, as they have the Tensor cores utilized for said technology. Just think, you could get twice the frames at half the cost at some point in the future when developers adopt the technology.

DLSS IN THE OPEN WORLD

Now, what about more wide open spaces, you ask? Give the No Man’s Sky comparison a fair shake below:

If you’re going to be rendering everything at once in a larger space, needless to say, that’s going to be more than a little taxing on your hardware. DLSS helps optimize things so that you can still have those fantastic vistas without tanking your frame rate. You may feel that the difference doesn’t seem that massive visually, but that is the whole point; understanding that you are able to enjoy and drink in all the details without your hardware buckling under the immense strain.

The more the tech gets used, the more the algorithm can learn, the faster it’ll be able to improve and deliver the experiences you’ll come to see as standard. It really can be just boiled down to having as low a requirement as possible, for the greatest potential output, and that’s why you’d want to get your foot in the door for NVIDIA DLSS for the long term result. Get yourself one of those GeForce RTX 20 or RTX 30 series and you could see yourself set for quite a long time.