Tuesday, March 17th 2026
NVIDIA CEO Claims Gamers are "Completely Wrong" About DLSS 5
At GTC 2026, NVIDIA announced its next-generation DLSS installment, version 5. However, after the community expressed significant backlash over the goals of DLSS 5, NVIDIA CEO Jensen Huang addressed the criticism surrounding the upcoming technology. In a Q&A session, NVIDIA provided Tom's Hardware with a response from a gamer's perspective about the objectives being pursued, as much of the negative feedback stemmed from the perceived visual downgrade in Resident Evil Requiem with the enhancements of DLSS 5. "Well, first of all, they're completely wrong. The reason for that is, as I have explained very carefully, DLSS 5 fuses the controllability of geometry, textures, and every aspect of the game with generative AI." What the CEO of the world's largest company is trying to convey is the controllability offered by DLSS 5 and how it provides developers with an option to enhance visuals.
For example, much of the criticism focuses on how DLSS 5 alters the visual definition of various in-game elements beyond the original developer's intention. Jensen Huang explains that DLSS 5 is not merely a post-processing tool for games but rather a system that gives developers generative control at the geometry level, not just a filter applied on top of the graphics processing pipeline. Huang noted that developers have complete control over what the technology does from the very beginning. DLSS 5 uses the game's color and motion vectors for each frame, and the model aligns with what the game developer originally intended for the image. Even the extent to which DLSS 5 enhances visual fidelity can be controlled, and it is entirely up to the developer to decide. "It's not post-processing, it's not post-processing at the frame level, it's generative control at the geometry level," iterated Huang.Gamers 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 its new visual fidelity enhancements. As NVIDIA positions DLSS 5 as a supplement to traditional technologies, no process or technology is dependent on it. Therefore, any major AAA game that receives an update to support DLSS 5 can also run normally without it. Players can still default to DLSS 4.5 for upscaling and visual enhancements, while DLSS 5 remains an option for those who want neural rendering to handle much of the visual heavy lifting.
Source:
Tom's Hardware
For example, much of the criticism focuses on how DLSS 5 alters the visual definition of various in-game elements beyond the original developer's intention. Jensen Huang explains that DLSS 5 is not merely a post-processing tool for games but rather a system that gives developers generative control at the geometry level, not just a filter applied on top of the graphics processing pipeline. Huang noted that developers have complete control over what the technology does from the very beginning. DLSS 5 uses the game's color and motion vectors for each frame, and the model aligns with what the game developer originally intended for the image. Even the extent to which DLSS 5 enhances visual fidelity can be controlled, and it is entirely up to the developer to decide. "It's not post-processing, it's not post-processing at the frame level, it's generative control at the geometry level," iterated Huang.Gamers 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 its new visual fidelity enhancements. As NVIDIA positions DLSS 5 as a supplement to traditional technologies, no process or technology is dependent on it. Therefore, any major AAA game that receives an update to support DLSS 5 can also run normally without it. Players can still default to DLSS 4.5 for upscaling and visual enhancements, while DLSS 5 remains an option for those who want neural rendering to handle much of the visual heavy lifting.


179 Comments on NVIDIA CEO Claims Gamers are "Completely Wrong" About DLSS 5
I think they should have had the release closer to their announcement; testing would have prevented the widespread backlash from a couple of seconds of video.
I can see RTX 6000 gpu's who are barely better than Blackwell in Raster/RT but can run this well happening now.
I mean if somebody doesn't have a PC and they want to build one that can run this what is it gonna cost. Geforce now FTW I guess, with that premium 6080 subscription lol.
They definitely got to start somewhere but this seems like more of a tech demo that some developers who are likely heavily compensated will include in their games.
There is also the real possibly of this making games cross the line into uncanny valley territory which in stills can look awesome but as soon as the characters start moving and talking fall apart because our brain knows what an actual real person looks like, moves like, and acts like.
You use your own local dedicated GPU for underlying geometry calculation, and then you have your live Geforce Now like streaming of end result filter on top of that - calculated by a server a million times more powerful than the fastest GPU. It will be economical, as calculating locally of course isn't, it will offer much better visuals than even the best GPU ever could, and for the special top customers they could offer tiers that could run 8K and above with all bells and whistles, offering a larger portion of server power, for a price of course.
And it will be completely locked in, of course. Subscription based, naturally.
Personally I hope I am wrong but that is honestly the first thing that popped into my head when I watched the trailer.
They still make 4 billion from gaming which is chump change comparted to their overall quarterly earnings but still a ton of money, my guess is a 6060 ultra DLSS5 edition is gonna cost 800 but it will do this better than a 5090.....
I'm just not going to blindly think something is going to be good for gaming either until developers/Nvidia prove it to be so.
I will be very impressed if Nvidia get's this working ok on 16GB cards at 1440p. I got a 5080 I can barrow anytime and test it on but I doubt it's going to be a good experience.
As a non-GeForce customer news for this technology doesn't really impact me.
Maybe there's a shred of DLSS support for my Quadro but I don't know/care.
I jump to an RTX card in the future and this technology doesn't impact the decision.
The more nVidia you buy, the more you'r dependent on their hardware cause investors will allways say "Make it big, make it look pretty, use the latest technology. The customer will buy our product and when it runs like dogshit, they will buy the latest hardware."
There is no competition in the hardware market if the customer is tied to your ecosystem by proprietary hardware and software.
Just repeat the phrase often enough and people will belive it. Upscaled and artificial sharpened images are better then native.
1. Nvidia has contradicting itself non stop about what the technology actually is and how it work. Never before a Nvidia technology has been presented in such a messy way and it doesn't aid that much of the verbose use to describe it, is imprecise at best.
2. The technology show some of the problem we have with image to image diffusion (image to image AI generation). Those limitations are inherent to that technology and not something that can really be fixed only mitigated at best (with LORA for exemple). This is a problem, because 1. Nvidia keeps telling us, DLSS5 isn't an AI filter (image to image diffusion) and 2. This will seriously limit the potential of DLSS5 to fixe all the issues we can already observe and make the technology worse as input resolution diminue.
Sometimes when I hear Jensen Huan I remember guy which was marketing slap chap it sounds similar, Jensen just talks a little slower, but for Me the level of marketing is on par.
Second thing is that Big can steal, small goes to jail, and big have been caught stealing work to train it's models :( and have no consequences.
Nvidia should separate the filter between the lighting one (like we see on the environment and far away character model) and a character/actor "enhancement" one and let gamers choose wich one they want to activate.
I'm sure Nvidia would very much like you to buy an AI coprocessor for DLSS though. :laugh:
For now it's low access and low impact. But I dread the idea of everyone going "this is a great idea, think of the money we'll save on designers!" and every game ending up with as much AI-generated slop graphics as possible.
Obviously that is the worst case scenario but it has happened before. Nvidia pushed ray-tracing, everyone implemented it or added support for it. Nvidia pushed upscaling, we now have FSR and XESS alongside DLSS.
"Just don't use it" applies right up until it's not optional anymore.
www.gamespot.com/articles/square-enix-reveals-stunning-final-fantasy-pc-tech/1100-6427022/