Nvidia’s 3000 series of GPUs are geared up with many standout features that set them apart as top-tier options for gaming and productivity. But this guide is going to touch on a lesser-discussed and more complex feature of these cards that is a bit more complex than simply having more cores, a higher VRAM cache, or less power consumption. DLSS is a newer form of anti-aliasing with a few extra bells and whistles. This technology brings a lot to the table when compared to older GPUs and systems.
For a more general overview of anti-aliasing (AA) and its varied forms, you can check out our guide for it right here. Some of the basics of AA will be covered here, but this overview will focus mainly on DLSS and how it improves the experience in modern titles.
DLSS is an acronym that stands for “Deep Learning Supersampling.” To understand what DLSS is compared to regular anti-aliasing, it’s best to take a look at the “Supersampling” bit of the equation. Supersampling is achieved through a method called downsampling. The game is rendered at a higher resolution and then shrunk down to fit the resolution of your display. This gets rid of rough edges and smooths artifacts that may be more visible at lower resolutions.
For example, Supersampling Anti-Aliasing (SSAA) could render a game in 4K UHD and then downsample it to a 2K display for a clearer experience. It’s one of the more common and most effective forms of anti-aliasing.
Nvidia improved on this already effective method using machine learning. That’s where the “Deep Learning” part comes in. But if SSAA is so effective, why would Nvidia want to improve it? Well, SSAA is very demanding on a system, as it takes a lot of resources to downsample effectively. Due to the resource demand, some systems don’t get much benefit from it due to the drop in frame rates.
Nvidia’s DLSS uses artificial intelligence (AI) to produce similar, if not better, results than SSAA without the hefty computing load. Nvidia uses an improved version of their previous-generation tensor cores to make this possible. Put simply, these cores are capable of performing the complex tasks associated with AI machine learning. 2000 series cards did have tensor cores of their own, but this new generation vastly improves on the technology. This means faster speeds and makes DLSS possible.
How Does DLSS Work?
As the name suggests, DLSS begins with supersampling. From there, an algorithm analyzes the rendered game and trains itself to do that same sampling but without using as many resources. This machine learning allows the GPU to produce images/frames that appear to be at higher resolutions. Essentially, it can make 1440p look like 4K and 1080p look like 1440p without using the resources required to natively output to those resolutions. It is a complicated process. To put it simply, the AI remembers certain aspects of the game and can render those remembered or logged assets faster and more accurately as time goes on. DLSS gets stronger as it gets used.
Regular AA can be fraught with bugs, artifacts, and inconsistencies. DLSS is geared at getting rid of those inconsistencies through learning and adapting over time. Nvidia purports that this technology will continue to grow more advanced as the years pass, enabling it to outperform past forms of AA. Right now, it isn’t without its fair share of bugs. However, constant optimizations have led to incredible improvements in recent years.
Some staggering claims are being thrown around. DLSS has always been able to garner more frames when compared to more traditional forms of AA. But those improvements have been relatively minor with 5-15% improvements. Recently, Nvidia has claimed vast improvements with upwards of 50% more frames in titles when compared to other forms of AA. These improvements are especially significant in lower resolutions like 1080p.
The Future of DLSS
DLSS gets better over time. Nvidia claims that this technology has a bright future ahead, especially as resolutions get more advanced. DLSS makes premium resolutions accessible at a much lower price point than displaying those same resolutions natively. 8K is still a ways off from being a reasonable option for the average consumer. However, those who have 4K displays will soon be able to use DLSS to achieve an experience similar to that of an 8K display without the hefty price tag.
Well, at least not the price tag of an 8K display. You’ll still need a powerful GPU like an RTX 3090 to handle that task. And DLSS can take your experience to the next level with improved frames while not sacrificing visual fidelity.
Since this technology is so new, there aren’t too many games that support DLSS yet. As the list grows, so, too, will the performance of DLSS as a whole. The list only has 50 or so titles on it right now, which includes games like Death Stranding and Cyberpunk 2077. With future Nvidia cards slated to have this technology, it’s safe to expect that list of compatible games to expand greatly over the next few years.