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
Add your own comment

153 Comments on NVIDIA Shows DLSS 5 Progress and Technical Details at SIGGRAPH 2026

#51
wolf
Better Than Native
ZoneDymothat implies that they cant change the minds though.
They'll be open to it if presented with better arguments and less toxic discussion.
ChomiqI mean, you can literally see the effect with the character's skin in ON/OFF comparison, why would I bother myself with a 50 minute keynote at a conference dedicated to 3D artists?
Perhaps to fully inform yourself for the thread literally about that content that, you're participating in.
Posted on Reply
#52
Chane
I still think this tech needs to be under some branding other than DLSS. Could have put it under the RTX Remix branding or like... Nvidia NeuralWorks, remember HairWorks everybody?
Posted on Reply
#53
TheDeeGee
Fake resolution, fake frames and soon fake art.

What a time to be alive.

I'm glad to have experienced actual raw hardware graphics progression in the late 90s and early 2000's.
Posted on Reply
#54
MTS-C
tsurugi

Um, you should know that what Reshade can do is very limited, as it is completely post-processing, with effects far inferior to DLSS5. Reshade RTGI is improving the effects of old games, with not much it can do, let alone surpassing modern games. When old games were created, relevant technologies were not popular, which is why there was room for improvement, whereas now RT has become the graphics standard for DX12, making these restricted Reshade RTGI completely unnecessary.

DLSS5 its at post-processing, but does the work of the rendering pipeline, and traditionally achieving the effects of DLSS5 could not be processed in real-time at all, as AI deeply processes the lighting, material rendering, textures, and subsurface reflections, which is different from simple Reshade effects.

With TAA, 1440p only offers 480p clarity, until DLSS appeared, allowing 1440p to reach 4k clarity, and the latest DLSS model can achieve near 4k using just 1080p. Then with the launch of MFG, the display refresh rate rather than the graphics card became the limit. FG added 5ms of latency, although the latency was completely offset by DLSS and Reflex and is far better than the latency provided by consoles, NVIDIA is just too evil.
Visually these 2 images aren't any more representative of doing more than Reshade than say what's going on in this video.

The amount of visual processing you can do with Reshade doesn't make DLSS5 in the shots from this thread look worth the argument.

DLSS does not make 1440p have the same visual quality as 4k native, sorry but that's cope.
Posted on Reply
#55
Onasi
TheDeeGeeFake resolution, fake frames and soon fake art.

What a time to be alive.

I'm glad to have experienced actual raw hardware graphics progression in the late 90s and early 2000's.
I am experiencing actual whiplash. So you want advancements that aren’t fake… and so you are not satisfied with the current GPU era… which is all about pushing RT and PT… which are FUNDAMENTALLY the least “fake” that 3D graphics can get. You DO realize that those “fake” resolutions and frames, originally at least, were meant to actually facilitate said push in terms of making the performance level required actually feasible, correct? Like, oh, all the tricks and fakery of the… uh… late 90s and early 00s… meant to facilitate real time lighting, texture filtering, polygonal environments, model density and other things.
Shiiiiiiet, it’s like all the graphix in my vidya were fake all along and the race is about finding new fakery… makes ya think. Or not. NV bad or whatever.
Posted on Reply
#56
wolf
Better Than Native
OnasiI am experiencing actual whiplash.… makes ya think. Or not. NV bad or whatever.
I think many commenters did not watch the video outlining the progress and what this latest iteration is, can do, and how it works.
Posted on Reply
#57
tfdsaf
How about keep proper game developers in the studios to make amazing games, get rid of the politically correct human resources eating up the money, fire all of the woke, no talent "developers" that add "woke" in the game and hire proper developers that create amazing stories, great characters, solid stories and pay them good money to keep them, rather than firing everyone and replacing it with AI slop that steals copyrighted images and videos in order to make similar copies of it?

I mean i don't see how this is good in any way shape or form. It is still changing the face of the characters to look like another person, even though its not so different now, it is still making it ultra realistic and brighter which might not be the artists intent and is probably going to have dumb CEO's fire their artists for this shit to create bad AI slop games that all look exactly the same!
Posted on Reply
#59
tsurugi
MTS-CVisually these 2 images aren't any more representative of doing more than Reshade than say what's going on in this video.

The amount of visual processing you can do with Reshade doesn't make DLSS5 in the shots from this thread look worth the argument.
I gave the modern game AC Shadows, while you gave Halo, an older game—that’s exactly why reshade can make difference, which validates my point. The most significant improvement comes from RTGI, but screen-space RTGI has flaws, and modern games have DX12 RT standards available, so Reshade RTGI is not needed.

Despite the improvements, halo reshade still looks like an old game, with nothing visually special and a huge gap compared to modern games, so I don't know what you are talking about. Are you really trying to beat DLSS5 with this visual quality?

Look at how DLSS5 changes the graphics, and see how DLSS5 alters light, object textures, and reflections to dramatically boost fidelity. reshade can only handle traditional post-processing, which are old hand-tuned algorithms we are all too familiar with, and it is precisely their effects cannot be improve any more that forced Nvidia to introduce AI RT/PT/DLSS5. If you dislike vanilla color grading and want to add sharpen, different AO, or cheap bloom, go ahead, but this simple post-processing barely improves on the original graphics (although adjusting unsatisfactory tone, sharpen, and AO to your liking is also an improvement) and can never compare to DLSS5.


I recommend putting my 3 DLSS5 on/off screenshots into Nvidia ICAT single screen mode for a comparison, , while reshade only provides some basic post-processing effects.
MTS-CDLSS does not make 1440p have the same visual quality as 4k native, sorry but that's cope.
If you really think so. Even at 4k, modern games must use TAA to avoid aliasing, which results in blurriness, and any other AA has noticeable aliasing unless you only appreciate static images. Only DLAA is better than DLSS , but DLAA is also evil AI.
Posted on Reply
#60
Chomiq
SnoutBaronProbably to understand more about the thing you're moaning about, perhaps. Also the DLSS 5 section is only around 25'ish minutes long.
wolfThey'll be open to it if presented with better arguments and less toxic discussion.

Perhaps to fully inform yourself for the thread literally about that content that, you're participating in.
I don't need that, the article provides enough detail. I see no use for this other than to use AI to piggyback on features that normally would come through raw GPU performance upgrades. Nvidia has hit the wall starting from 30-series onwards and major breakthroughs don't come from raw GPU performance upgrades but through features like DLSS which piggyback on AI/ML tech to cheat their way to better visual quality and performance. And yeah, AMD is guilty of the same, just on smaller scale since they're yet to hit 600W at the wall just from the GPU alone.

Nvidia did exactly what I expected after the initial AI slop backlash, they toned things down to a point where they seem almost irrelevant and you need to be told what to look for to notice actual differences. Complete 180 from the AI slop filter in RE:9. Sure, it looks good in still screenshots but falls apart as soon as you that character moving their head and you see artifacting from the AI filter on top of character's face.

But hey, I need to watch entire SIGGRAPH keynote to be able to discuss this.

Posted on Reply
#61
Bigshrimp
Slop is still slop, even it looks less like slop. I still don't see the need for this, as this is just more glorified gimmicks to sell to the slop eaters. They will do anything to sell GPUs, ANYTHING...
Posted on Reply
#62
MTS-C
tsurugiI gave the modern game AC Shadows, while you gave Halo, an older game—that’s exactly why reshade can make difference, which validates my point. The most significant improvement comes from RTGI, but screen-space RTGI has flaws, and modern games have DX12 RT standards available, so Reshade RTGI is not needed.

Despite the improvements, halo reshade still looks like an old game, with nothing visually special and a huge gap compared to modern games, so I don't know what you are talking about. Are you really trying to beat DLSS5 with this visual quality?

Look at how DLSS5 changes the graphics, and see how DLSS5 alters light, object textures, and reflections to dramatically boost fidelity. reshade can only handle traditional post-processing, which are old hand-tuned algorithms we are all too familiar with, and it is precisely their effects cannot be improve any more that forced Nvidia to introduce AI RT/PT/DLSS5. If you dislike vanilla color grading and want to add sharpen, different AO, or cheap bloom, go ahead, but this simple post-processing barely improves on the original graphics (although adjusting unsatisfactory tone, sharpen, and AO to your liking is also an improvement) and can never compare to DLSS5.


I recommend putting my 3 DLSS5 on/off screenshots into Nvidia ICAT single screen mode for a comparison, , while reshade only provides some basic post-processing effects.

If you really think so. Even at 4k, modern games must use TAA to avoid aliasing, which results in blurriness, and any other AA has noticeable aliasing unless you only appreciate static images. Only DLAA is better than DLSS , but DLAA is also evil AI.
We're talking about noticeable visual differences in a before or after. Whether a game is old or not, the before and after is the final product. Look at the two images you just provided and tell me you would tell a difference without watching static images. Your argument should be reversed. The comparison between the two is night and day on Halo vs nobody would notice a single difference in motion in the DLSS5 images. Your original argument is that reshade is 'just' a post processing and can't do anything remotely as good as DLSS5. Your recent examples vs mine tell a different story. In fact, older games would be the best representation TO show DLSS5 improvements because those games don't have high visual fidelity at all and goes back to the core argument people had with DLSS5 to begin with. How much are we willing to change in original art intention? Take away the AI slop faces from their first announcement and you have 2 images that look incredibly similar and wouldn't be noticed at all in motions during normal gameplay.

As for the second half of the DLSS convo, nobody ever talks about the noise that comes with using RT or DLSS/FSR. NOISE. When you're trying to claim 1440p can have the same clarity as 4k you can't forget the fact that these techs add noise in motion. Denoising is never perfect. There's bluring and noise which you don't get with native. If you need a different kind of TAA to replace TAA to still wind up with a less shit version maybe try something else. Problem is by the time the 3000 series came out people were complacent with it and decided to forget all the hate we had for all of this from the 2000 series introduction.
Posted on Reply
#63
JustBenching
Chomiq, AMD is guilty of the same, just on smaller scale since they're yet to hit 600W at the wall just from the GPU alone.
Because they don't have gpus equivalent in performance to the 5090. The nvidia cards are as fast or faster than amds fastest consume less power than the amd cards, so what in the world is even your point?
Posted on Reply
#64
tsurugi
MTS-CWe're talking about noticeable visual differences in a before or after.
visual differences are hard to represent numerically, but I personally think the improvement from DLSS5's visual differentiation is no less than Halo with reshade enabled.
MTS-CWhether a game is old or not, the before and after is the final product. Look at the two images you just provided and tell me you would tell a difference without watching static images.
Of course, the difference in visual quality is very noticeable, especially during dynamic gameplay, where dlss5 fidelity is great for AAA games.
MTS-CYour argument should be reversed. The comparison between the two is night and day on Halo vs nobody would notice a single difference in motion in the DLSS5 images. Your original argument is that reshade is 'just' a post processing and can't do anything remotely as good as DLSS5. Your recent examples vs mine tell a different story. In fact, older games would be the best representation TO show DLSS5 improvements because those games don't have high visual fidelity at all and goes back to the core argument people had with DLSS5 to begin with. How much are we willing to change in original art intention? Take away the AI slop faces from their first announcement and you have 2 images that look incredibly similar and wouldn't be noticed at all in motions during normal gameplay.
Old games, like starfield, yes. But I suspect truly old games will need at least RTX Remix support for modern graphics APIs to use DLSS5.
It's your game, so you can play it however you want, and graphical overhauls using mods make it even better—sometimes developers do this themselves, remastering and reselling.
MTS-CAs for the second half of the DLSS convo, nobody ever talks about the noise that comes with using RT or DLSS/FSR. NOISE. When you're trying to claim 1440p can have the same clarity as 4k you can't forget the fact that these techs add noise in motion. Denoising is never perfect. There's bluring and noise which you don't get with native. If you need a different kind of TAA to replace TAA to still wind up with a less shit version maybe try something else. Problem is by the time the 3000 series came out people were complacent with it and decided to forget all the hate we had for all of this from the 2000 series introduction.
So you need to use Nvidia AI -based Ray Reconstruction denoiser to eliminate noise. To run smoothly without noise in PT, you must use the full suite of features provided by Nvidia, including MFG, DLSS, and RR. These features either eliminate image artifacts or boost performance/smoothness. Besides acting as a denoiser, RR also participates in ray reconstruction to improve image quality. All of this is necessary for achieving real-time ray tracing.
AI is very suitable for denoising. I have provided Intel Open Image Denoise examples for comparison.
Posted on Reply
#65
Chomiq
JustBenchingBecause they don't have gpus equivalent in performance to the 5090. The nvidia cards are as fast or faster than amds fastest consume less power than the amd cards, so what in the world is even your point?
The point is that we don't get progress in raw RT and raster performance and instead we get crutches like DLSS.
Posted on Reply
#66
tsurugi
Doctor Whohow AMD is the dominant CPU player and Intel is crawling from behind, copying AMD at every chance they get.
Just like AMD tried to learn Intel E-core's state-of-art cluster decoder, nice try but failed, Zen 5 ST only has 4-wide, and another 4-wide cannot work without SMT.
Just like Intel has long had advanced EMIB packaging, which makes AMD's substrate-based Chiplet look like no big deal, and Foveros, which Intel has used for years, AMD still hasn't caught up with and still uses worse interconnect technology in Zen 6.
Just like AMD tries to copy Intel's product naming, Intel calls it i7, AMD calls it R7, Intel renames to Ultra, AMD immediately renames to AI, terrified of straying from Intel's naming scheme.
Just like AMD, even in slides comparing against Intel's products, only dares to use Intel's other products instead of their own.
Posted on Reply
#67
MTS-C
tsurugivisual differences are hard to represent numerically, but I personally think the improvement from DLSS5's visual differentiation is no less than Halo with reshade enabled.

Of course, the difference in visual quality is very noticeable, especially during dynamic gameplay, where dlss5 fidelity is great for AAA games.

Old games, like starfield, yes. But I suspect truly old games will need at least RTX Remix support for modern graphics APIs to use DLSS5.
It's your game, so you can play it however you want, and graphical overhauls using mods make it even better—sometimes developers do this themselves, remastering and reselling.

So you need to use Nvidia AI -based Ray Reconstruction denoiser to eliminate noise. To run smoothly without noise in PT, you must use the full suite of features provided by Nvidia, including MFG, DLSS, and RR. These features either eliminate image artifacts or boost performance/smoothness. Besides acting as a denoiser, RR also participates in ray reconstruction to improve image quality. All of this is necessary for achieving real-time ray tracing.
AI is very suitable for denoising. I have provided Intel Open Image Denoise examples for comparison.
Your entire post reads nothing more like an empty PR skit. You can't possibly sit there and say that the noticeable difference between the Halo demonstration and the DLSS5 images you provided are either on par or the Halo one is LESS. Like the lighting differences alone stand out greatly in MOTION in the video on Halo where as your DLSS5 images you have to stare and nitpick to see it. For instance this quote here reads off like a political redirect at best "Of course, the difference in visual quality is very noticeable, especially during dynamic gameplay, where dlss5 fidelity is great for AAA games." because no, it does not. Look at those images you provided and tell me with a straight face you'd notice the difference in motion. You can't. It's either you're being purposely obtuse or you are generating your replies with AI.

Obviously ray-reconstruction is used to negate some of the noise but it's not perfect and doesn't take it all away. When you add in the upscaler it replaces some of the noise with blur, and MFG adds latency to the mix. This isn't it chief. And again, using still pics to try and prove a point isn't going to work because the noise is most noticeable in motion. That's like putting 2 images in a post saying see you can't tell the difference between 120fps and 60fps.
Posted on Reply
#68
AleXXX666
GeofrancisIt's just more AI slop, it won't get used to enhance anything, it will just get used to pump out cheap sloppy games. just like the ai video generators.
that's the era we live in, take it or not lol
I don't give a F about "electric cars" but I liked to be driven in few Teslas. (but NOT their Chinese crap-clones lmfao)
Posted on Reply
#69
evernessince
I couldn't find it in the article but according to other sources this version now uses other game information now like albedos, which is definitely a step in the right direction.

I'm looking at the images and can see it still has the tendency to add that front facing lighting look. Compare the side of the books in image 1 and 2. I assume Nvidia specifically chose front facing lighting to try and hide the model's natural tendency to do so, even at lower denoise strengths. There's more detail in the watermelon but it's not really detail that should be there. It looks too rough for a watermelon. The concrete wall in image 2 is oversharpened but that might just be DLSS in general.

I noticed a lack of focal depth as well, wondering how the model might handle that or other effects.

This DLSS 5 off image struck me as extremely uncanny.

It appears to be made specifically to make the DLSS 5 look good

Layering additional creative controls, especially ones that don't coincide with the existing ones, I can imagine will be rather annoying for devs. Another set of knobs to tune and one's that don't correspond with the existing ones you have. A dev might be concerned that lighting in a particular scene doesn't look as it should so they dig through the lighting and shaders, only to find it was something DLSS 5 was doing. Just a potential scenario.

I also take issue with the claim that DLSS 5 is adding subsurface scattering at 22:00 of the video. As far as I'm aware, they don't show controls for specific lighting techniques. You can't tick just subsurface scattering as far as they've demonstrated. You get the lighting the AI thinks the scene should look like, with the parameters Nvidia gives you pulling the model in certain direction (emphasis on pulling as the model's internal understanding still takes precedence). Even if they did train a model just on subsurface scattering, you'd still get bleed from the training data. That's just a flaw of current AI models.

It seems they've attempted to address the lack of artistic consistency (to some success) but realistically many of the challenges they needed to overcome were fundemental to current AI models. The performance cost will also be a question of course as well. If this is being used to enhance details, it has to be kept to a very lower performance impact. I question the usefulness for enhancing lighting until they completely get rid of the AI look. I guess it looks better than the PT images they used as an example but those just look horridly uncanny.
Posted on Reply
#70
tsurugi
MTS-CYour entire post reads nothing more like an empty PR skit. You can't possibly sit there and say that the noticeable difference between the Halo demonstration and the DLSS5 images you provided are either on par or the Halo one is LESS. Like the lighting differences alone stand out greatly in MOTION in the video on Halo where as your DLSS5 images you have to stare and nitpick to see it. For instance this quote here reads off like a political redirect at best "Of course, the difference in visual quality is very noticeable, especially during dynamic gameplay, where dlss5 fidelity is great for AAA games." because no, it does not. Look at those images you provided and tell me with a straight face you'd notice the difference in motion. You can't. It's either you're being purposely obtuse or you are generating your replies with AI.
What I said is what I think, dlss5 indeed shows significant improvement. This is my screenshot of DLSS5 real-time rendering in Nvidia's latest video, which is dynamic, and you can watch Nvidia's video yourself. It is not that reshade has obvious color changes while DLSS5 does not, meaning the change is small. Less is more. Images processed by DLSS5 look very realistic and textured, possessing the photographic aesthetic of cinematic visuals and well-set lighting, with a noticeable boost in visual fidelity. In my view, this enhances immersion quite well. I believe many people will feel the same way, though I cannot guarantee everyone shares the same mindset, so it is completely fine if you do not feel that way.
MTS-CObviously ray-reconstruction is used to negate some of the noise but it's not perfect and doesn't take it all away. When you add in the upscaler it replaces some of the noise with blur, and MFG adds latency to the mix. This isn't it chief. And again, using still pics to try and prove a point isn't going to work because the noise is most noticeable in motion. That's like putting 2 images in a post saying see you can't tell the difference between 120fps and 60fps.
DLSS 4.5 RR can solve your problem. coming in August
Posted on Reply
#71
Trunks0
tsurugi

Um, you should know that what Reshade can do is very limited, as it is completely post-processing, with effects far inferior to DLSS5. Reshade RTGI is improving the effects of old games, with not much it can do, let alone surpassing modern games. When old games were created, relevant technologies were not popular, which is why there was room for improvement, whereas now RT has become the graphics standard for DX12, making these restricted Reshade RTGI completely unnecessary.

DLSS5 its at post-processing, but does the work of the rendering pipeline, and traditionally achieving the effects of DLSS5 could not be processed in real-time at all, as AI deeply processes the lighting, material rendering, textures, and subsurface reflections, which is different from simple Reshade effects.

With TAA, 1440p only offers 480p clarity, until DLSS appeared, allowing 1440p to reach 4k clarity, and the latest DLSS model can achieve near 4k using just 1080p. Then with the launch of MFG, the display refresh rate rather than the graphics card became the limit. FG added 5ms of latency, although the latency was completely offset by DLSS and Reflex and is far better than the latency provided by consoles, NVIDIA is just too evil.
The more complex Reshades can do allot, especially when they start using plug-ins to go beyond just typical post processing. Which is why I said its kinda crazy what reshade can really do. You could very very easily get extremely close to most of what nVidia was showing(not all, but most). Which makes allot of sense, as DLSS5 is mostly working as a post process itself... hence my comment that it looked like an "AI Reshade", because I would expect an AI based post processing method to be able to do exactly what they demonstrated.

We are still kinda at the start of people grasping what can be done in real-time with AI. So I 100% expect we will see reshades use AI and manage to do almost everything shown. I also would 1000% expect engines like UE, Frostbite, Anvil, Snowdrop and etc to implement features like this that work universally, have way more control and allow for way more flexability than what nVidia has shown.
Posted on Reply
#72
Fatloader
Eh, it looks less like garbage than it used to. I'm looking forward to trying it out and then inevitably turning it off due to garbage performance on my 5070.

Whatever happened to baked lighting for static scenes?
Posted on Reply
#73
tsurugi
Trunks0The more complex Reshades can do allot, especially when they start using plug-ins to go beyond just typical post processing. Which is why I said its kinda crazy what reshade can really do. You could very very easily get extremely close to most of what nVidia was showing(not all, but most). Which makes allot of sense, as DLSS5 is mostly working as a post process itself... hence my comment that it looked like an "AI Reshade", because I would expect an AI based post processing method to be able to do exactly what they demonstrated.

We are still kinda at the start of people grasping what can be done in real-time with AI. So I 100% expect we will see reshades use AI and manage to do almost everything shown. I also would 1000% expect engines like UE, Frostbite, Anvil, Snowdrop and etc to implement features like this that work universally, have way more control and allow for way more flexability than what nVidia has shown.
Yes, Reshade might build in AI features, which would be really fun.
Posted on Reply
#74
evernessince
tsurugi

Um, you should know that what Reshade can do is very limited, as it is completely post-processing, with effects far inferior to DLSS5. Reshade RTGI is improving the effects of old games, with not much it can do, let alone surpassing modern games. When old games were created, relevant technologies were not popular, which is why there was room for improvement, whereas now RT has become the graphics standard for DX12, making these restricted Reshade RTGI completely unnecessary.
Reshade can gain access to things like depth buffer and there are full reshade ray tracing shader available.

You are definitely underselling it.
tsurugiDLSS5 its at post-processing, but does the work of the rendering pipeline, and traditionally achieving the effects of DLSS5 could not be processed in real-time at all, as AI deeply processes the lighting, material rendering, textures, and subsurface reflections, which is different from simple Reshade effects.
It's not that said effects aren't traditionally achievable. RT, PT, subsurface scattering all are in realtime without the need for DLSS 5.

It's the "quality" of those effects. AI has a fixed cost for high level of lighting fidelity. The problem is that it carries of large variety of other issues with it. Right now those lighting effects are bundled, you can't choose just subsurface scattering. AI has bleed issues as well, where it'll pull outputs towards it's common understanding. It's why you still see the front facing lighting effect, because AI is trained on images with that kind of lighting.

The problem of not being able to unbundle specific effects is inherent to AI. The neurons are all connected and one input can pull on many neurons. I have to assume as well that in order to keep performance reasonable, understanding is limited. It's not going to be on the level of a model dedicated to identifying things visually. Originally it probably wasn't identifying things at all but merely coloring pixels as that's all it had beside motion vectors. Them feeding it more game data will absolutely help. The point of a model like this isn't to deeply process the data, it's to get it through the neural network as quickly as possible with the highest possible output aesthetic quality.
Posted on Reply
#75
BigMack70
I do not and will not ever want an AI slop filter over my games.
Posted on Reply
Add your own comment
Nov 19th, 2026 12:23 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts