Revolutionizing Graphics: A Deep Dive into DLSS Technology by NVIDIA (part one)

Estimated read time 7 min read

A few years back, NVIDIA introduced a series of graphics cards that came with a couple of rather groundbreaking ideas for the time. The first of the two was ray tracing, a technology that literally simulates the physics of light instead of relying on cheap tricks to achieve more lifelike illumination. NVIDIA was so confident in this endeavor that they even changed the name of their graphics cards from GTX to RTX in honor of Ray Tracing. However, they also introduced another, less flashy technology that garnered a bit more attention over time: DLSS, a technique aimed at making games run smoother using artificial intelligence. 

In 2024, and now six years later, while Ray Tracing undoubtedly still exists, the spotlight has gradually shifted towards DLSS, which has seen significant evolution and proven to be the more practical of the two technologies. Numerous companies have either created or are in the process of developing their own versions, indicating that DLSS is shaping up to be the present and future of gaming. 

As part of the ongoing revolution in artificial intelligence, which is permeating various industries, including gaming, let’s delve into how DLSS works. The reality behind DLSS is far more impressive than one might initially think. 

We are currently witnessing a revolution in the field of artificial intelligence. Major companies like Amazon, Google, Meta, and many others have been utilizing AI for a myriad of purposes for years. Now, this technology is reaching everyday users. You, sitting at home, can engage in conversations with chatbots about your emotional problems, generate photos with MidJourney, create 3D scenes with Nerf, or even simulate someone else in a video. These are all feats we’ve explored on this channel. However, gamers envision even more possibilities. 

Imagine infusing artificial intelligence into a video game. If you’re a true gamer, your first thought upon trying out chatbots might have been, “It would be awesome to have this in a video game.” Picture non-playable characters engaging in endless conversations with distinct personalities, dynamically generated missions, and other such innovations. While this vision may come to fruition eventually, we already have a successful case of using AI in video games, and it’s been quite a remarkable revolution: Deep Learning Super Sampling (DLSS). 

But first, let’s understand how video games operate on a computer. When we play a game, the graphics card is responsible for generating the images we see on screen in real-time. Essentially, while you play, a chip is continuously rendering millions of pixels to create full-screen images, giving us the sensation of animation and movement. 

However, these images are entirely static; they are essentially photos within the video game, similar to those in videos and movies. While resolution may not matter much in games like Dead Cells or Undertale, it significantly enhances the experience in others like Resident Evil, providing richer, more detailed imagery. Hence, in recent years, many users have been transitioning from using Full HD monitors (1920×1080 pixels) to Quad HD (2560×1440 pixels) or even 4K (3840×2160 pixels) monitors. While higher resolutions offer more detail, they also require the graphics card to render more pixels, resulting in decreased frame rates. 

For instance, upgrading from Full HD to 4K translates to four times the workload for the graphics card. To illustrate, if you get 100 frames per second (fps) at 1080p, you might only get around 70 or 60 fps at 4K. This drop occurs despite the graphics card and the processor both being involved in the rendering process. More pixels per image mean more work, resulting in fewer frames per second. And frame rate is crucial in gaming. While movies and videos often run at 24 or 30 frames per second and still look good due to motion blur and the brain’s perception of motion, gaming is different. Each frame in a game is generated at the moment you see it, making them static. Therefore, to achieve natural fluidity, games require a higher frame rate. 

Ideally, for smooth and enjoyable gameplay, frame rates between 120 and 240 Hz (120-240 fps) are desirable. While such high frame rates aren’t necessarily required, they do enhance the gaming experience. In fact, modern laptops come equipped with displays boasting refresh rates of up to 240 or even 360 Hz. However, there’s a trade-off between resolution and frame rate. Lowering the resolution makes it easier for the graphics card to calculate frames, resulting in higher frame rates. Conversely, increasing the resolution reduces frame rates, as each frame requires more rendering work. 

So, where does DLSS come into play? NVIDIA’s idea was to leverage artificial intelligence to assist in this process. The question was, at which stage could AI help us extract more power from the same graphics card? The initial concept was to decrease resolution by 50% or 70% and then reconstruct the image using AI to make it appear as if it were at 4K resolution. By reducing the resolution, we could achieve higher frame rates. Therefore, a game running at 60 fps with many effects and realistic graphics at 4K could potentially be boosted to 90 or 100 fps using this technique. And the idea is indeed brilliant. 

Thus, DLSS 1.0 was born. DLSS 2.0 aimed to enhance graphical effects, improving certain aspects, as we’ll explore later. DLSS 3.0, on the other hand, focuses on increasing frame rates once again, employing artificial intelligence. But before delving deeper, let’s pause for a moment to precisely explain what we mean by artificial intelligence, a distinction that many people fail to grasp fully. What exactly is artificial intelligence, and how does it differ from, say, NPC artificial intelligence? 

NPC artificial intelligence in video games is based on predefined algorithms programmed by humans. These algorithms dictate the NPC’s behavior, reactions, and routines under various circumstances. While the complexity may vary, NPC AI essentially follows a set of clear and concise instructions. On the other hand, when we refer to DLSS utilizing artificial intelligence, we’re talking about something entirely different. 

In the case of DLSS, artificial intelligence involves training a mathematical model to produce outputs based on given inputs. Rather than programming explicit instructions, we’re training a system to generate instructions autonomously by showing it examples of situations it might encounter and how it should ideally respond. Essentially, it’s a learning machine. However, contrary to popular belief, while this technology is complex, it’s not as magical as one might think. Technologies like ChatGPT and others are continually improving and learning, but they won’t suddenly become sentient beings. Let’s explore this further. 

NVIDIA has been fortunate in that graphics cards, in addition to rendering game graphics, are exceptionally adept at handling learning algorithms. Although graphics cards weren’t explicitly designed for this purpose, they outperform processors significantly in this regard. 

This fortuitous aspect has contributed to NVIDIA’s current golden age. Much of the world’s artificial intelligence runs on NVIDIA hardware, a testament to their investments over the years. They’ve developed libraries, utilities, the CUDA API, and collaborated closely with researchers, placing them in a formidable position in this field. It makes perfect sense for NVIDIA to venture into this realm, merging their two areas of expertise: gaming and artificial intelligence. 

One of the most significant recent advancements in artificial intelligence has been Convolutional Neural Networks (CNNs). 

As we conclude this exploration into the fascinating realm of DLSS technology, we’ve merely scratched the surface of its potential. But fear not, for our journey has only just begun. In the next chapter, prepare to delve even deeper into the inner workings of DLSS, uncovering its secrets and witnessing its transformative impact on gaming like never before. Get ready to embark on an exhilarating odyssey through the wonders of artificial intelligence and gaming innovation. Join us in the next chapter as we unveil the full extent of DLSS’s prowess and discover how it continues to revolutionize the gaming landscape. Don’t miss out on this thrilling continuation of our adventure – stay tuned for Chapter 2 and unlock the next level of gaming excellence! 

+ There are no comments

Add yours

Leave a Reply