Thursday, March 19th 2026
NVIDIA DLSS 5 Gets 84% Dislikes on YouTube as Backlash Grows
NVIDIA's latest DLSS 5 technology has faced a significant community backlash, with its approval rating dropping considerably. On the published YouTube video, NVIDIA's official DLSS 5 announcement has received an overwhelming 83.7% dislikes, with only 16.3% likes. This is a substantial negative rating, with 16,107 likes and 82,515 dislikes (and counting) on a video with 1,527,915 views at the time of writing. Other videos published by the NVIDIA GeForce YouTube channel have also recorded surprisingly low approval from the community. The Resident Evil Requiem video only scored a 14.9% positive rating, while Starfield had an 18.2% positive ratio of likes to dislikes. Other demos such as Hogwarts Legacy and EA Sports FC saw positive ratings of 18.7% and 14.5%, respectively. The best rating is now exclusive to a tech demo, not even a real game, which is the Zorah Unreal Tech Demo with a 37% positive ratio.
Gamers' reactions are shifting negatively towards the technology, while NVIDIA CEO Jensen Huang famously noted that gamers are "completely wrong" because these games offer massive programmability and controllability in how DLSS 5 is applied, keeping the artistic intent intact. However, according to game developers from both Capcom and Ubisoft who spoke to Insider Gaming, while the individual studios may have been involved in marketing DLSS 5, the teams who worked on them were just as surprised by the results as the rest of the gaming community. A Ubisoft developer is quoted as saying, "We found out at the same time as the public," while developers at Capcom expressed similar sentiments, stating that it was surprising to see Capcom, which has generally been protective of its IPs when it comes to AI involvement, getting involved in the marketing for DLSS 5. Furthermore, the Capcom developers expressed concern about how DLSS 5 might change Capcom's approach to generative AI and its role in game development.We are still waiting to see how NVIDIA handles the rollout of DLSS 5, which is expected to happen later this year. Gamers will have the freedom to choose how they want to experience scenes with photorealistic lighting and materials. DLSS 5 can be completely turned off for a regular gaming session, or it can be set to the maximum for enhanced visual fidelity. Since NVIDIA positions DLSS 5 as a supplement to traditional technologies, no process or technology is dependent on it.
Source:
VGC
Gamers' reactions are shifting negatively towards the technology, while NVIDIA CEO Jensen Huang famously noted that gamers are "completely wrong" because these games offer massive programmability and controllability in how DLSS 5 is applied, keeping the artistic intent intact. However, according to game developers from both Capcom and Ubisoft who spoke to Insider Gaming, while the individual studios may have been involved in marketing DLSS 5, the teams who worked on them were just as surprised by the results as the rest of the gaming community. A Ubisoft developer is quoted as saying, "We found out at the same time as the public," while developers at Capcom expressed similar sentiments, stating that it was surprising to see Capcom, which has generally been protective of its IPs when it comes to AI involvement, getting involved in the marketing for DLSS 5. Furthermore, the Capcom developers expressed concern about how DLSS 5 might change Capcom's approach to generative AI and its role in game development.We are still waiting to see how NVIDIA handles the rollout of DLSS 5, which is expected to happen later this year. Gamers will have the freedom to choose how they want to experience scenes with photorealistic lighting and materials. DLSS 5 can be completely turned off for a regular gaming session, or it can be set to the maximum for enhanced visual fidelity. Since NVIDIA positions DLSS 5 as a supplement to traditional technologies, no process or technology is dependent on it.

132 Comments on NVIDIA DLSS 5 Gets 84% Dislikes on YouTube as Backlash Grows
DLSS 5 in a nutshell.
This probably was posted already but just in case not
That reason is games are now being arbitrarily forced to have this upscaler crap to even be playable which should never have happened.
OR we just going to skip over the incremental hardware gains that everyone is complaining about.
RT will go the same physX went DEAD & GONE & simplified crap from too much processing needs.
One of the biggest challenges with RT lighting, is the fidelity of the lighting simulations. Even with combining multiple frames to gather more detail, the ray traced lighting is still a tiny fraction of the viewport resolution, and thus it still can't present highly detailed light interactions, and sampling for too long will mean laggy lighting effects where they trail actions in the scene for too long and become distracting.
A better solution would have been to apply infer from the low number of light samples. Since the RT simulations will take care of lighting even from elements off-screen, it might be enough to generate a high res light map, rather than the current method of doing around 240-360p worth of rays, and then sampling multiple frames and applying a ton of denoising and then spatial interpolation as needed. It would look better than the current method of additional tone mapping and generating new faces.
Overall, instead of trying to apply generative elements across the entire scene which is very taxing to do in realtime, it could possibly be less demanding to use it just for the existing RT elements using the low res sample that may be 360p and have it guide a generative AI light map based on assets within the scene.
The whole thing confirms my long standing suspicion that most people who talk about tech on forums or in comment sections haven't got the first clue of what they are talking about. As well as the one that those who should know better and often do it for a living are willing to bend the facts to the narrative that excites the masses.
There's no point in engaging in "discussion" with anybody about this because we have already reached the death-threats and water-is-dry stage, so I guess all that's left is posting this brilliant chart:
Anything with less than 200 Tensor Cores wont make 1080p @30 fps
330+ Tensor cores to go over 60 fps
And 24+ GB to not suffer with stutter badly.
Definitely forget anything bellow $1000 ($2000 if thing go south).
PS: Steam HW Survey says rtx 5070Ti, 5080, 5090 all together make < 4%.
What we gon do? Play games without all of this bullshit.
Its so visually realistic "Do you think they got any shipment of cheese wheels in at the corner market today, Townton is so beautiful this time of day"
NVFP4 is another rtx 50xx thing.
rtx 4090 may be mighty, but sorry. Go buy something newer.
"The more you buy the more you save."
Enjoy it.Neural rendering + NVFP4 ... sounds as clear enough message it wont work on anything older than rtx 50xx series card.That is a good one.
Thank for posting.
FSR and DLSS are not turn-key features, and they actually require a game developer to dedicate time and resources towards implementing it well. Furthermore, I am sure we have all seen a game implement FSR or DLSS poorly and it has a ton of artifacts, only to be later fixed in a patch. If a turn DLSS 5 into something that only 5090 equivalent cards can run, then it will fail to command any meaningful developer time since those resources will be better focused on catering to the super majority of users.
Think how little attention AMD's Redstone featureset is getting relative to older FSR versions when it comes to how fast games are officially implementing the full featureset.
Also compare that to DLSS 3.5 and newer. You will see that the gap is only growing, simply put a feature that less then 1% of gamers can use, will not be a high priority for a game studio that is already going through crunch to meet a release date.
In my opinion, DLSS/RTX were always a gimmick at their base to introduce artificial difficulty in rendering to make sure older hardware became obsolete faster to get you to buy the newer model of card, to make you rely on these technologies to get the best experience with a mix of what the artists intended. There hasn't been a huge shift since we got Ambient Occlusion/Upscaling combined with sharpening techniques a fairly long time ago.
To add...
If for any reason these technologies from Nvidia are trying to replace previous rendering techniques, I don't believe there's a good outcome from that... but that's just me.
"They" already had the majority of the graphics card market and trying their best to make sure it stays that way. Then again, soon some of us might not even own hardware anymore.
I played Borderlands 2, Crysis 2/3. Skyrim kitted out or had a shader suite popped on to games if I wanted to change the way I perceive what is on the monitor (yes it was still FPS costly at the time depending on what post-processing was stuck on). The only nice thing that came out of the new tech for me was RTX remix which lets people do what SweetFX/ReShade did to an extent. Which is nice to have when devs don't have the source code or there isn't a source port of the original game to add in or change things that some people would prefer.
By the time I played last of what Unreal Engine 4 had to offer. That was the point where "traditional" rendering techniques had peaked. I was fine with playing most titles (engine agnostic) without messing with stuff like applying additional hair physics, remember Hairworks? That died quick and I think that was because it didn't add enough impact to rendering.
I doubt that some of us forget what happened while we were still keeping our 900/1000 series cards at the turning point.
I believe some tech enthusiasts that are more critical of the difference DLSS/RTX makes, realized that something isn't right here.
Let's call it DLBS from now on, because BS is all it is.
Just imagine if all of these efforts were put into making the most amazing drivers ever.
Buggy, lifeless, bad executed beauty.
If you will demand better gameplay, diverse story and surprising ideas than AI might give you just that.
But going only visuals is doomed to more and more AI upscaler makeup.
Lol what ???