What Is Nvidia DLSS for Gaming?

Nvidia DLSS has been round because the RTX 2080 dropped again in 2018, however whereas it began as a means to make use of machine studying to upscale video games, it is grown to be a lot greater than that.

Now, 8 years after the Tensor Core that powers DLSS first debuted in Nvidia’s Volta GPU structure, DLSS will upscale your video games, generate fully new frames, and, when DLSS 5 comes out later this yr, will even re-draw every body in your sport. Some of those options are extra divisive than others, however it’s onerous to argue that DLSS is not one of the vital necessary GPU software program suites shortly, and is a significant a part of why Nvidia graphics playing cards are so good.

DLSS Upscaling

DLSS ostensibly stands for “Deep Learning Super Sampling,” and that’s precisely what it did at first. The complete thought was to render the sport at a decrease decision, after which use an AI algorithm, educated on Nvidia tremendous computer systems and operating on Tensor Cores, to precisely upscale to the next decision.

For most video games, DLSS is offered in one in every of 4 presets, every altering the scaling issue of the sport.

• Ultra Performance: 33%
• Performance: 50%
• Balanced: 58%
• Quality: 66.7%

Obviously, the upper you go up, the higher the picture high quality goes to be, however that can conversely have an effect on your body fee. As a common rule of thumb, I’d suggest ‘Performance’ for 4K, ‘Balanced’ for 1440p, and ‘Quality’ for 1080p gaming. I personally flip DLSS on every time I can, however that’s as a result of the algorithm is so good today that I can’t inform the distinction except I’m actively searching for it. That wasn’t at all times the case, although.

In the start, this was tough. The first iteration of DLSS upscaling was noisy and had a ton of points with artifacts and common picture softness. Coupled with the truth that it solely ran in a few video games, and it was onerous to take Nvidia’s upscaling tremendous critically. However, with DLSS 2.0, launched in 2020, Nvidia improved the algorithm by taking movement vector knowledge from the sport engine into consideration, making it in a position to extra precisely generate pixels that have been true to the movement on the display screen.

DLSS 2 additionally shifted Nvidia’s upscaling algorithm from needing to be educated on every sport individually, to an algorithm that might be utilized to many alternative video games, assuming builders labored the DLSS recordsdata into their sport. This made adoption skyrocket, and we began to see a ton of video games beginning to help the know-how.

While the upscaling tech received just a little higher over the following couple of years, the following largest soar got here with DLSS 4, which debuted with the RTX 5080 in 2025. This moved the upscaling algorithm from a CNN (convolutional neural community) to a transformer mannequin, which drastically improved accuracy.

This replace labored on each RTX graphics card, and you may even force-enable it within the driver software program, permitting it to run in principally any sport that already supported DLSS 2 or 3. The solely draw back was that the mannequin was a tad heavier than the older CNN mannequin, so it may reduce the efficiency enhance you bought from enabling DLSS – significantly on older RTX 3000 and 2000 graphics playing cards.

Then, in 2026, Nvidia launched a bizarre little mid-generation replace, with DLSS 4.5. This modified each upscaling and body era – which I’ll get into later – however for upscaling it simply made the transformer mannequin a bit extra correct if you set it to “Performance” or “Ultra Performance” modes. The solely drawback was that it elevated the efficiency footprint on older playing cards – so I’d solely suggest enabling it with an RTX 5060 or newer graphics card.

Frame Generation

DLSS upscaling has change into fairly widespread over the previous couple of years, however with DLSS 3 and the RTX 4090, Nvidia went from utilizing AI to generate additional pixels, to producing complete frames. Frame gen is extremely easy in idea: It makes use of an algorithm to generate additional frames after which inserts them between really rendered frames within the render queue. However, that is after all a bit extra difficult than it sounds.

By their very nature, video video games are unpredictable, so Nvidia needed to discover a technique to generate the additional frames that wouldn’t flip video games into a large number of hallucinations and artifacts. And, similar to the upscaling bit, body era was tough at first.

Rather than simply being free efficiency like Nvidia initially claimed, body era elevated latency, relatively than reducing it. One of the primary functions of taking part in video games at the next body fee is to decrease the delay between no matter you’re doing – clicking a mouse, transferring your character, or no matter – and that motion being mirrored on the display screen. Because body era isn’t really making the sport run quicker, it will probably solely ever add latency, as a result of the algorithm wants a few milliseconds to generate every body.

Nvidia offsets this improve in latency with Reflex, a system that primarily zeroes out the render queue and has the CPU sync with the graphics card, in order that frames aren’t simply sitting there ready for the GPU to render them.

Reflex has been out for years, and it actually does enhance latency fairly a bit, however it’s essential to cease body era from being straight up horrible. With body era, Nvidia principally added a brand new render queue that exists on the GPU itself, taking a rendered body, holding it till a brand new body is generated, after which pacing them out to maintain the movement easy.

This can be why body gen doesn’t precisely double your body fee, despite the fact that it’s producing a brand new body off of each one despatched by the sport. And it’s additionally why it hits latency a bit. In my expertise, the latency will usually go from, for instance, 30 ms to 40 ms in Cyberpunk 2077 when body era is enabled.

The added latency isn’t splendid, however it’s not sufficient to be noticeable, particularly in single participant video games the place you’re not attempting to compete.

As controversial as body era was when it launched with the RTX 4080, although, Nvidia upped the multiplier with DLSS 4 in 2025, including a 3x and 4x body gen choice. Now, RTX 50 collection graphics playing cards may generate as much as three frames per rendered body, theoretically boosting body charges by 4x. Again, that quantity doesn’t precisely work out in idea, however it does vastly enhance efficiency.

Generating all of these rendered frames may have been a nightmare, however Nvidia labored an AMP core, or AI Management Processor, into its Blackwell (RTX 50 collection) graphics playing cards. This little chip works as a form of taskmaster for the GPU and takes over body pacing, which was usually dealt with by the CPU beforehand. Because that work is now dealt with on the identical die because the GPU, there may be much less latency when scheduling out the frames, which is why multi-frame era is ready to work.

What’s extra spectacular, although, is how little the additional generated frames influence latency past the preliminary latency influence of 2x body gen. In the identical Cyberpunk part I discussed earlier, turning body gen to 4x solely bumps the PC latency as much as 43ms, which is barely extra. And whereas including latency in any respect is a draw back, the upside is that it’s a lot simpler to make use of a 4K 240Hz monitor and get your cash’s price.

But 4x body gen wasn’t sufficient for Nvidia. At CES 2026, the corporate introduced DLSS 4.5, a form of half-generation refresh of DLSS 4. With it, 6x body gen is now doable, however the extra thrilling a part of the equation is Dynamic Multi Frame Generation. When enabled, this may have the GPU, um, dynamically change the body gen multiplier, to maintain your body fee as near your monitor’s refresh fee as doable.

Think of it as a form of spicy V-Sync, however as an alternative of limiting body fee to your refresh fee, it simply generates additional frames to maintain your monitor absolutely fed. With this replace, Nvidia additionally up to date the AI mannequin that body era runs on. Creatively referred to as “Model B” within the Nvidia App, it’s a lot better at dealing with UI components, which may mess up in some video games when body gen is lively.

Nvidia appears to be heading in a route of regularly rising the quantity of frames that DLSS goes to generate, and who is aware of the place it’s going to determine to hit the brakes. But irrespective of how excessive that multiplier will get, simply needless to say this function positively isn’t for anyone.

If you’re nonetheless rocking a 60Hz show, body era just isn’t going to do something for you. And in the event you’re struggling to hit 50-60 fps within the first place, body era goes to be a laggy, artifact-riddled mess.

Instead, body era is finest for people who find themselves already getting 50-60 fps on a high-refresh show, the place it’ll make video games look a lot smoother. Even then, although, the added latency signifies that lots of aggressive avid gamers ought to in all probability simply ignore it.

The Future of DLSS

While each DLSS upscaling and body era use AI to enhance efficiency, they don’t meaningfully change the way in which the sport itself seems. While there is likely to be little errors right here and there, the ultimate product is extraordinarily devoted to what the sport is outputting. But that is likely to be altering later this yr when DLSS 5 launches.

All we’ve actually seen from DLSS 5 to date is a brief demo on-stage at GTC 2026 (Graphics Technology Conference), so there’s not so much to go off of. But, primarily based on what was there, this new algorithm appears to massively influence the aesthetics of a sport, significantly with regards to character fashions.

Nvidia claims that the mannequin is grounded within the geometry and “scene semantics” of the sport. But when the mannequin is taking the sport’s last output together with movement vector knowledge to generate a picture that’s overlaid on prime, it certain seems prefer it’s altering the aesthetics of the sport.

In idea, which means the higher the underlying sport seems, the higher DLSS 5 will probably be at producing a last picture, however as a result of it doesn’t take knowledge from the sport engine into consideration, it’s very doable that it’s going to mess up. Though, Nvidia will probably patch it as much as make it look higher over the following half-a-year earlier than it’s out there to obtain.

But past the influence to a sport’s aesthetic, it stays to be seen whether or not or not it’s going to really enhance efficiency. After all, because the starting, one of many guiding lights of DLSS was maximising efficiency whereas preserving picture high quality. The early mannequin of DLSS 5 wanted two RTX 5090s to run. And, certain, Team Green will discover a technique to optimize it to run on a single graphics card, however solely time will inform if it’ll be one thing that avid gamers activate. If it simply adjustments the way in which the sport seems however with worse efficiency – I don’t suppose lots of people are going to decide on to make use of it.