Loading local time… change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts

Tuesday, July 21st 2026

NVIDIA Shows DLSS 5 Progress and Technical Details at SIGGRAPH 2026

by
AleksandarK
 Discuss (148 Comments)
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

Related News

Add your own comment

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

#1
Dr. Dro
I am quite glad Nvidia took all the criticism and mockery to heart, this is impressive. But I still expect extremely high system requirements, 5080+, perhaps even excluding the 4090 if it requires something that only Blackwell and Rubin can do.
Posted on Reply
#3
Onasi
It looks less like a shitty TikTok filter. So that’s good. I am still not sold on the idea fundamentally, but not looking like hot ass aesthetically is a nice first step. Now it actually would need to PERFORM in a reasonable manner, which I am not really sure about being feasible on anything besides flagship cards.
Posted on Reply
#4
Geofrancis
It'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.
Posted on Reply
#5
Frizz
I think we can bet that the most valuable company in the world has the best chance of making proper use of AI in comparison to the "slop" out there. So far the results look impressive.
Posted on Reply
#6
RejZoR
RecusWatch AMD fanboys attack this but suddenly it's must have feature when AMD has alternative. (FSR x8 frame generation)
Sorry, but this is the same shit that you can achieve with ReShade. On any graphic card. For free. Performance and obtaining it wise. Open both images in different tabs and alternate between them. It's just a not very impressive filter. The "original" DLSS5 was impressive on just how much crap they can smear on top of actual graphics using Ai models. This showcase, they are just showing some hardware accelerated ReShade filters and it's really not impressive at all. But hey, it'll be an option I guess.

Also no one is asking for multiframe generation because input latency is absolute garbage, even on NVIDIA. What I would ask for are further improvements to the AFMF. That shit is awesome. Zero perceivable input latency, 2x perceived framerate. If they can figure out how to better exclude crosshairs and HUD from it and it would be even more amazing.
Posted on Reply
#7
john_
I don't care how people reacted, or react, I KNOW that Nvidia will push this version of DLSS and people will love it. We all know it.
Posted on Reply
#8
JustBenching
john_I don't care how people reacted, or react, I KNOW that Nvidia will push this version of DLSS and people will love it. We all know it.
If it's good, of course people will love it. If it isn't, people won't. DLSS 1 as an example. Stop going down the anti nvidia rabbit hole.
RejZoRAFMF. That shit is awesome. Zero perceivable input latency\
Have you tried it? It adds more latency than multi frame generation. Way more. Easily noticeable even when your base framerate is high. Tried it on SOTR at ~80 base fps.
Posted on Reply
#9
RejZoR
JustBenchingIf it's good, of course people will love it. If it isn't, people won't. DLSS 1 as an example. Stop going down the anti nvidia rabbit hole.


Have you tried it? It adds more latency than multi frame generation. Way more. Easily noticeable even when your base framerate is high. Tried it on SOTR at ~80 base fps.
No it doesn't. I've been using it since launch of RX 9070 XT. I can't sense it on mouse at all and I'm sensitive as hell to that. FSR3 FrameGen in games has been horrible every time I tried it and has only mildly improved recently with FSR4 FrameGen. Only in Marvel Rivals which also implemented AMD Anti-Lag 2 which probably helps it. I can still feel it a bit and it's annoying. AFMF on the other hand, apart from slight artifacting on crosshair, basically perfection. To claim FSR FrameGen is better is even further weird because that was just never the case for me.
Posted on Reply
#10
SnoutBaron
RejZoRAlso no one is asking for multiframe generation because input latency is absolute garbage, even on NVIDIA.
You must speak for a lot of people, I'm surprised. I don't recall saying you could speak for me though. I like MFG and I have a feeling, and now I don't want to speak for everyone like you do, but I think there will be quite a bit of people who would like MFG on AMD cards.
Posted on Reply
#11
Dr. Dro
RejZoRSorry, but this is the same shit that you can achieve with ReShade. On any graphic card. For free. Performance and obtaining it wise. Open both images in different tabs and alternate between them. It's just a not very impressive filter. The "original" DLSS5 was impressive on just how much crap they can smear on top of actual graphics using Ai models. This showcase, they are just showing some hardware accelerated ReShade filters and it's really not impressive at all. But hey, it'll be an option I guess.

Also no one is asking for multiframe generation because input latency is absolute garbage, even on NVIDIA. What I would ask for are further improvements to the AFMF. That shit is awesome. Zero perceivable input latency, 2x perceived framerate. If they can figure out how to better exclude crosshairs and HUD from it and it would be even more amazing.
...no. ReShade can't do anything like this, the most you can get out of ReShade is a CAS (aka what FSR 1 used to be). They're doing this because people whined about artistic intent and that it was slopification of graphics (and they weren't wrong). Arguing for AFMF (and mind you your request will never happen, because the entire point of this feature is to output the framebuffer twice - latency increases accordingly and you are just ignoring that) while lashing out on DLSS-G is hilarious, though. I will keep both of them off...
Posted on Reply
#12
Chomiq
And now there's minimal difference between DLSS off and on so what's the point? You're just adding useless shit to the frametime pipeline.
Posted on Reply
#13
Dr. Dro
ChomiqAnd now there's minimal difference between DLSS off and on so what's the point? You're just adding useless shit to the frametime pipeline.
If it's light enough to be used as a detail enhancer and runs on midrange HW, it's a win on my book.
Posted on Reply
#14
Chomiq
Dr. DroIf it's light enough to be used as a detail enhancer and runs on midrange HW, it's a win on my book.
However, they generally fit and run fine on a single GPU without any problems.
This bit doesn't sound promising.
Posted on Reply
#15
Dr. Dro
ChomiqThis bit doesn't sound promising.
Yeah, I'm not expecting this to run on anything below a 5080... maybe 4080
Posted on Reply
#16
SnoutBaron
ChomiqAnd now there's minimal difference between DLSS off and on so what's the point? You're just adding useless shit to the frametime pipeline.
At 21:40-22:00 of the presentation, which I'm sure you've watched before giving an opinion, they give an example of using DLSS 5 to fake 'realistic' subsurface scattering in a raytraced scene, which they claim is more performant than actually applying subsurface scene-wide. Now as you say, it's a small difference...but game graphics today...is made up of years of small graphical improvements compounded together.
Posted on Reply
#17
john_
JustBenchingIf it's good, of course people will love it. If it isn't, people won't. DLSS 1 as an example. Stop going down the anti nvidia rabbit hole.
Stop trying to ruin every thread when you feel that your favorite company needs your defense. It doesn't.
Posted on Reply
#18
rafipc2025
Why not to work on dlss 4 and dlss 4.5 to make it perfect? More fps for less lag and better quality.
Why push of this new crap for new cards i bet, if it will hit fps a lot, what's even point of releasing it?
Thing is that many people cant afford of this new shit overpriced gpu's and even new platform with ddr5.
Dlss was made to make fps more right? Or im missing something?
Next 6th gen nvidia gpu will give us even less probably then 5% of performance increase now i bet.
Posted on Reply
#19
Hecate91
It doesn't matter if DLSSlop 5 is good or turns games into an Instagram filter, people will say they love it regardless, especially the diehard fans. And plenty of reviewers will have "preview content" praising how great the AI slop is.
Posted on Reply
#20
Dr. Dro
rafipc2025Why not to work on dlss 4 and dlss 4.5 to make it perfect? More fps for less lag and better quality.
Why push of this new crap for new cards i bet, if it will hit fps a lot, what's even point of releasing it?
Thing is that many people cant afford of this new shit overpriced gpu's and even new platform with ddr5.
Dlss was made to make fps more right? Or im missing something?
Next 6th gen nvidia gpu will give us even less probably then 5% of performance increase now i bet.
You're not missing just something, you're missing a whole lot, but I don't blame you. The marketing is awful. This is not an upscaler and does not replace what is being marketed as DLSS 4.5, but a whole new component: a generative AI image enhancer.

As of "4.5", DLSS now features several components:
Super Resolution, which upscales the image and corrects imperfections (nvngx_dlss.dll) - DLSS 2 introduced first render presets, with new versions just being new presets available on new DLLs
Ray Reconstruction, which is an advanced AI RT denoiser (nvngx_dlssd.dll)
Frame Generation, which extrapolates frames to multiply the frame rate (nvngx_dlssg.dll)
Posted on Reply
#21
sbacc
Make game use bad AA and TAA so you can advertise DLSS

Make game use bad lighting and reflection so you can advertise RTX

Make game use bad material rendering, making everything plastic-ish so you can advertise DLSS5

What is the next step ? Probably make game logic/physic/NPC AI worse so you can advertise a proprietary magic band aid to the rescue.

Edit : in 10 years the "Nvidia OFF" version of game will look and play so bad, old game will have to be delisted and potentially "remaster" to keep the illusion...oh wait they're already doing it !
Posted on Reply
#22
Onasi
rafipc2025Why not to work on dlss 4 and dlss 4.5 to make it perfect? More fps for less lag and better quality.
…but they literally just recently updated those modes with new, better models. Like, what are you even talking about.
rafipc2025Dlss was made to make fps more right? Or im missing something?
No, that wasn’t the only reason.
Posted on Reply
#23
john_
rafipc2025Why not to work on dlss 4 and dlss 4.5 to make it perfect? More fps for less lag and better quality.
Why push of this new crap for new cards i bet, if it will hit fps a lot, what's even point of releasing it?
Thing is that many people cant afford of this new shit overpriced gpu's and even new platform with ddr5.
Dlss was made to make fps more right? Or im missing something?
Next 6th gen nvidia gpu will give us even less probably then 5% of performance increase now i bet.
Probably Nvidia feels that DLSS in it's current form is as good as it can get. That's why it offered the latest version on older cards also. They probably ended with it. We might see a 4.6, or 4.7 something but that's it.

DLSS 5 on the other hand can offer Nvidia a visual advantage over the competition. Nvidia always tries to have a proprietary technology that will be offering them a clear advantage over the competition. That's why they keep those technologies proprietary and never offer them to others. PhysX, GameWorks libraries, HairWorks, GSync, DLSS 1-4. Now they will push DLSS 5 and believe me tech sites slowly will start talking about the most important evolution of computer graphics the last 20 years, because we will have a true transformation of what we see on screen, not just better lighting effects as it is the case with raytracing for example. And considering that DLSS 5 needs newer hardware and stronger hardware, it is the way that any company will want to go, so it can sell more GPUS, more expensive GPUs.
AMD will follow, Intel will follow (not necessarily in that order, I do expect Intel to beat AMD in this one because AMD is focused on AI and server CPUs), I don't know if Qualcomm and Apple can follow. I am expecting Mediatek being in cooperation with Nvidia to gain a significant advantage over Qualcomm in the next 3-5 years.
Posted on Reply
#24
MarkoB88
Dr. DroI am quite glad Nvidia took all the criticism and mockery to heart, this is impressive. But I still expect extremely high system requirements, 5080+, perhaps even excluding the 4090 if it requires something that only Blackwell and Rubin can do.
As the launch is expected this fall, the company will provide general system requirements and further details
Posted on Reply
#25
Dr. Dro
MarkoB88As the launch is expected this fall, the company will provide general system requirements and further details
I read the post, mate. But even as someone who's in the safest zone possible here (I own an RTX 5090), if this needs such a card to run, it'll be a bad experience even for us.
Posted on Reply
Add your own comment
Loading local time… change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts