Tuesday, July 21st 2026
NVIDIA Shows DLSS 5 Progress and Technical Details at SIGGRAPH 2026
NVIDIA has once again showcased its DLSS 5 neural rendering at the SIGGRAPH 2026 event, currently taking place in Los Angeles. Despite the initial negative reaction from gamers, NVIDIA is continuing to develop the technology, offering improvements and a new demonstration of how DLSS 5 is intended to function. One of the most interesting aspects is that game developers can choose from three DLSS 5 models, each providing a different level of structural intensity, global illumination, texture detail, and more. These models, labeled Model A, Model B, and Model C, can be tailored to specific scenes or character needs, allowing developers to mix and match according to their requirements. Each model carries a different parameter count, meaning that some are more compute and memory-intensive than others. However, they generally fit and run fine on a single GPU without any problems.
Another noteworthy feature is the preservation of artistic intent, which was heavily criticized in the original DLSS 5 demo. Each individual object, scene, and game setting can be customized according to the developer's vision. This is one of the challenges NVIDIA faced with DLSS 5, as maintaining visual identity, object semantics, and character poses, along with output lighting semantics, motion, and art direction, is crucial. Within these strict parameters, DLSS 5 applies material photorealism to enhance visual appeal. Below you can even see the customization of the applied model, where there are adjustable parameters that can be done per scene and per character.The second challenge that the company faced is temporal streaming. Traditional generative AI video generators process in chunks, reasoning across a frame and generating a batch of frames at once. However, real-time rendering requires immediate generation, frame by frame. Since DLSS 5 uses motion vectors from the game engine, it can generate additional textures instantly, without delay. This results in no shimmer or image corruption, as DLSS 5 know exactly where to attach itself and perform its texture generation instantly.The final challenge for DLSS 5 is speed. To render a 4K frame with 8.3 million pixels at over 60 FPS, NVIDIA achieves this in less than 16 ms, with most of the processing time consumed by the game engine rather than the DLSS 5 model. The model is so efficient that users will not notice any delay during gameplay, and it operates on a single GPU. While NVIDIA highlights VRAM efficiency, it remains unclear whether 4K gaming will be limited to high VRAM GPUs or something else. As the launch is expected this fall, the company will provide general system requirements and further details, but the improvements compared to the original demo seem promising for now.
Source:
NVIDIA at SIGGRAPH 2026
Another noteworthy feature is the preservation of artistic intent, which was heavily criticized in the original DLSS 5 demo. Each individual object, scene, and game setting can be customized according to the developer's vision. This is one of the challenges NVIDIA faced with DLSS 5, as maintaining visual identity, object semantics, and character poses, along with output lighting semantics, motion, and art direction, is crucial. Within these strict parameters, DLSS 5 applies material photorealism to enhance visual appeal. Below you can even see the customization of the applied model, where there are adjustable parameters that can be done per scene and per character.The second challenge that the company faced is temporal streaming. Traditional generative AI video generators process in chunks, reasoning across a frame and generating a batch of frames at once. However, real-time rendering requires immediate generation, frame by frame. Since DLSS 5 uses motion vectors from the game engine, it can generate additional textures instantly, without delay. This results in no shimmer or image corruption, as DLSS 5 know exactly where to attach itself and perform its texture generation instantly.The final challenge for DLSS 5 is speed. To render a 4K frame with 8.3 million pixels at over 60 FPS, NVIDIA achieves this in less than 16 ms, with most of the processing time consumed by the game engine rather than the DLSS 5 model. The model is so efficient that users will not notice any delay during gameplay, and it operates on a single GPU. While NVIDIA highlights VRAM efficiency, it remains unclear whether 4K gaming will be limited to high VRAM GPUs or something else. As the launch is expected this fall, the company will provide general system requirements and further details, but the improvements compared to the original demo seem promising for now.











153 Comments on NVIDIA Shows DLSS 5 Progress and Technical Details at SIGGRAPH 2026
This is the worst case from gta v:
Like I said before, during the car acceleration there is a smear under the bumper, you can;t miss it.
Cyberpunk 2077 is the same withe cars. (red engine)
Oblivion remastered is the same (ue5). When your weapon accelerates and decelerates it leaves after images and they are huge. You can't miss them.
Dlss 5 I suspect will do the same but the whole frame. The colors will shift and move slightly. Changing and leaving trails like an eye trick:
(but actually moving)
Here is an example of shimmering at native that dlss fixes, the video is from my pc
So as I've said, that's a very dishonest argument cause we are supposed to pretend that only upscaling causes this which is a lie.
Also, the introduction of DLSS 2.0 is the problem you described. They keep getting away with undersampling effects that would look broken without TAA and not optimizing stuff. And Nvidia can keep selling their overpriced, underperformed hardware since they intend to use software tricks like DLSS to replace raw performance
There has to be a balance. DLSS or frame generation is not that balance. Because of the majority market share, tech goes in all sorts of directions for Nvidia. So you're right, no one *has* to use DLSS, but that industry steering impacts game development overall. where poorly optimized games get away with it just because of DLSS making up for it.I agree completely. Good point. One of my main criticisms of DLSS 5 or rrr, the AI slop, is that I don't like the fact it changes character appearances and artistic intent, among several other things. DLSS 5 actually re-imagines and generates the final appearance of objects and lighting.
I'll add: "Give me an upscaler that can improve performance "without" ruining a game's artistic intent and without using fake generated frames, which makes games look fake."Couldn't agree more, regardless of company.
For example, gta v is unplayable for me with dlss (up to 4). I will test 4.5 more to see how it looks, first impressions are good for the first time ever.
Same thing with many other games be it dlss, depth of filed, games bugs etc...
Dlss 5 creates similar concerns about image clarity. That is the point.
It's tough because to some degree, they're damned if they do and damned if they don't.
Visual advances? They cost too much performance! AI slop! Nobody asked for RT! Nobody asked for PT! Nobody asked for realism, et cetera and so on. Some people demand more of the same performance, some people want newer better at the same perofmance. I'd argue we're getting both.
DLSS is genuinely a boon imo because it pushed forward what AA and TAA was doing and is giving people more tools (way more by extension of other vendors and companies also competing, ala FSR, XeSS etc) to get ever-better results from sub-native rendering, and of course even native AA through the best algorithms available, which has been happening since well before dlss 1.0
I really don't see the direct link asserted that upscaling is stagnating visuals. It's progression in parallel to, and to the betterment of, increased polygon counts, geometry density, texture quality, lighting enhancement, realism enhancement, increased visual fluidity and so on.
The other guy said it best, DLSS is TAA. What does TAA do? It blurs aliased edges for the appearance of a smoother object in motion. If everybody is in agreement motion blur is terrible in game why would anybody be in agreement this is ok for a few extra FPS? If you want more FPS turn your settings down or reduce resolution like we've always done. DLSS and in retrospect RT has been one of the biggest brainwashing psyops in tech IMO.
Oh, and your opinion on die space “wasted” by Tensor cores is noted and discarded. They are not there for your vidya to run better, that’s just a fortunate byproduct.Oh, yeah yeah, totally. When I was seething at CDPR when playing The Witcher 2 because they dithered all shadows and alphas on vegetation to absolute shit that totally was NV and their DLSS… more than half a decade before it was actually released, but no matter. Or those undersampled shader and shadow effects in GTA IV. Or terrible undersampled screenspace effects in half of UE3 games. But nah, it’s obviously new and NV is to blame.
Since I'm biased I just asked my gf. It took her less than half a second, she said top looks way better. So I pressured her a bit more
That was genuinely our dialogue. So yeah, I guess she is also biased, even though she didn't know what she was looking at
Show her in motion, not stills. You've ignored that your entire time trying to convince yourself it looks better between myself and the other user you're arguing with. But I digress, continue licking the boot and being a part of the problem. If you like a blurry mess in motion by all means, the other user can continue arguing with a biased brick wall.
Both images are zoomed in btw the exact same amount, you can literally print screen and put it on paint, you can match one atop of another 1:1. So yeap, clearly im the one being biased :D
I beg everyone reading this go show these stills to a random person in your family and ask them which one is better. Don't tell them anything about dlss upscalers or whatever, just show them the images. Let's see whos actually biased here, I bet a 100k everybody will find top clearly, infinitely better.
EG1. Here, I went through the trouble of proving that both are equally zoomed in. I moved the top part of the guy (dlss) to the bottom (native). Exact same size...Accusing me of bias while thinking 2 identical pictures are zoomed differently. Just wow man
Tensor cores are an unfortunate byproduct of Nvidia tailoring its modern GPUs toward data centers instead of graphics rendering. They also take up 1/3 of die space per SM in the latest 5000 series, they should be replaced by all shading cores for pure raw performance instead.
At this age there's not even a whisper whether a multi-GPU could elevate this, the only thing I heard is Lossless Scaling being able to use secondary GPU to do the fake frames.
Righto. I'll move on and continue to not give a flying fuck about this
This reminds me strongly of the 'showcase mode' TVs have and use in the showroom. It is just there to grab your attention and make you buy a TV, at home you'll never use it at all. I'm genuinely wondering what's the actual point of this technology at all, as I have done since announcement. Nothing fundamentally changed - DLSS still warps color and balance of the picture no matter how you set it up.Don't forget about the epeen value of being able to afford such a fantastic technology.
Its a bit like driving a Ferrari. Looks nice, but in most situations also completely does not fit the use case, for example the first speed bump you meet, which will take off your skirts if you drive over it; and now your car produces artifacts :D. And then comes the question what kind of gamer are you: do you game to buy that ferrari, or do you game to drive fast. I reckon most do the latter; and that Ferrari only gets in the way of it.
RT > DLSS = lots of wasted capital, time and effort that could have gone into either cheaper GPUs or actual better game art. That's the long and short of it. These technologies never existed to make graphics better, they exist to upsell bottom tier content and GPUs. To get back to the car analogy, every GPU is a car and can drive just fine. RT/DLSS is the nitro button that makes it all go slower and faster at the same time and practically makes no difference in getting from A to B.
www.screenhub.com.au/wp-content/uploads/sites/4/2026/03/nvidia-dlss-rendering-system.jpg
i.ytimg.com/vi/nHrn_4ujjYo/sddefault.jpg
Or any Fallout / TES title... I mean artists definitely do manage to make us question 'the intent'...
But in the end, maybe that's artistic intent, we still talk about it, best marketing ever made.
Signed - 'The Fan'
I mean... DLSS this, m.f.ker :D
Am I taking crazy pills? Nvidia should follow AMD in this regard?
Did you read or watch the parts where the game developers are going to have complete control over how or indeed even if DLSS5 is implemented in their game and in exactly what granular quantities? Per scene, per cutscene, per level, per object type(s), per groupable objects/masks, with 0-100 sliders for intensity, with multiple model/presets, and it this is still a pre release showing. When the time comes and games ship with DLSS5 enhancing it, it will by necessity of the way the toolset works, be the artistic vision of the developers.
And even that to some degree is a statement I don't think is some indisputable ideal to hang our hats on. For example if a game ships with and enables by default, as per the developers artistic vision, motion blur, depth of field, chromatic aberration, film grain, lens flare, vingette and so on, do we bemoan people disabling any of those to ensure they are able to enjoy the game without distraction/frustration/nausea?