Wednesday, September 16th 2026

AMD Rumored To Have DLSS 5-Like Neural Rendering Coming With RDNA 5

AMD's next-gen RDNA 5 architecture might get its own neural rendering feature, something in the same ballpark as NVIDIA's DLSS 5. That's according to Kepler_L2, a well-known AMD leaker, though it's worth being clear that this is a single-source claim and should be treated as a rumor for now. Posting on the Anandtech forums, Kepler_L2 said AMD will have its own version of "Neural Lighting" comparable to DLSS 5 with its next-gen hardware. Asked whether this would be limited to RDNA 5, the leaker said it's probably going to be the case, since this kind of workload is apparently pretty heavy. That's about as far as the concrete part of the leak goes.

Kepler_L2 also shared thoughts on how AMD's approach should differ from NVIDIA's. He thinks the model should have direct access to raw game data like geometry and depth buffers instead of having to guess at scene details, plus splitting lighting and shading into separate models instead of one pipeline. It's worth stressing that this part is the leaker's own wishlist, and not a description of what AMD is actually building. AMD hasn't confirmed RDNA 5 by name yet, and launch estimates currently range from mid-2027 to early 2028 depending on who you ask. Therefore, even if this pans out, we're probably looking at something that's early in the development pipeline.
Source: Kepler_L2 on Anandtech Forums
Add your own comment

75 Comments on AMD Rumored To Have DLSS 5-Like Neural Rendering Coming With RDNA 5

#26
ZoneDymo
I mean...have fun I guess for those that want this fake crap
Posted on Reply
#27
TheinsanegamerN
DenverAMD and Nvidia have been working on neural rendering for the past few years, more or less at the same time, so that's not true.

- AMD developed HBM in collaboration with SK Hynix, perhaps they still regret to this day that they opened it up to the world instead of keeping it exclusive, like GDDR6X was between Nvidia and Micron.
AMD could have kept it to themselves, but with their execution it would have withered and died. (Oh hi vega).
Denver- Low-level APIs were pioneered by AMD. Vulkan exists because of Mantle, which was a positive development for the Linux community. Vulkan is just as good as a rendering API as it is as a compute API.
3DFX beat AMD to the low level API by 25 years.
Posted on Reply
#28
TumbleGeorge
Prima.Verainstead of coming themselves with something new and innovating?!?
Maybe tech creativity about hardware of all human engineers and scientists is has it completely exhausted itself, or do they believe they have tried everything that is applicable to the current ecosystem and there is nothing better that can be printed in the hardware? And for some reason they do not want to change the ecosystem from the ground up. Maybe they are just too lazy.
Posted on Reply
#29
harm9963
Plenty of Mods already doing this and only getting better .
Posted on Reply
#30
Darmok N Jalad
TheinsanegamerNAMD could have kept it to themselves, but with their execution it would have withered and died. (Oh hi vega).

3DFX beat AMD to the low level API by 25 years.
If you mean Glide, it was great, but poorly supported by developers. Quake II had Glide (and 3Dnow!), and performance skyrocketed, but no other games I played offered it. Then 3DFX hit hard times and sold their IP to NVIDIA, who has apparently never did anything with that tech.
Posted on Reply
#31
Easo
Hecate91Nothing but a rumor until AMD confirms this.
It'll be interesting if true,perhaps less sloppy AI slop if the AI has direct access, but I don't want AI slop in games no matter who it's from.
You have to be realistic - AMD cannot have no answer to DLSS 5.
Posted on Reply
#32
AusWolf
It'll be definitely RDNA 5 exclusive... Until they retroactively release a fallback version for RDNA 4 and 3. We know how things go these days.
Posted on Reply
#33
Hecate91
EasoYou have to be realistic - AMD cannot have no answer to DLSS 5.
Yeah AMD essentially has to, however both Nvidia and AMD following with these features are creating problems in search of solutions, graphics already peaked years ago, now AAA studios are putting graphics above all else. Neural rendering is too graphically intense for not much visual difference in many cases when it's toned down to not look completely like AI slop. Also while we've had ray tracing for 8 years and it still can't be done on lower mid range cards without lower resolution and adding in AI rendered frames.
Anything AMD creates themselves or could hold exclusive like HBM or Vulcan would get trashed by the press and social media, they have no choice but to follow what Nvidia is forcing on the market.
AusWolfIt'll be definitely RDNA 5 exclusive... Until they retroactively release a fallback version for RDNA 4 and 3. We know how things go these days.
New GPU releases are rumored to be delayed until 2028, it would be dumb to make this RDNA5 exclusive. But that is how it works anymore, companies restrict features and modders have to prove it's possible to port back features to previous cards.
Posted on Reply
#34
AusWolf
Neural this, neural that... It's all basically just matrix operations. Don't they run on common instruction sets? Why do they have to be brand exclusive?

At this rate, developers will have to make two entirely separate versions of the same game, one for Nvidia, one for AMD. It's incredibly dumb, imo.
Posted on Reply
#35
Darmok N Jalad
AusWolfNeural this, neural that... It's all basically just matrix operations. Don't they run on common instruction sets? Why do they have to be brand exclusive?

At this rate, developers will have to make two entirely separate versions of the same game, one for Nvidia, one for AMD. It's incredibly dumb, imo.
Yeah, and NVIDIA has been partnering/sponsoring development for some time. Remember all those “The Way it’s Meant to be Played” splash screens on games? NVIDIA was helping fund those games to get that level of sponsorship. Such partnerships are not without benefits, even long-term ones. It’s like Intel helping pay for OEM product advertising on TV. If Intel wanted to advertise itself, they certainly could, but it’s way better to make your brand inseparable from the product it’s in, and it has worked pretty well for decades. If there are competing APIs, who do you think is going to get the most attention when the developer resources get assigned?
Posted on Reply
#36
AusWolf
Darmok N JaladYeah, and NVIDIA has been partnering/sponsoring development for some time. Remember all those “The Way it’s Meant to be Played” splash screens on games? NVIDIA was helping fund those games to get that level of sponsorship. Such partnerships are not without benefits, even long-term ones. It’s like Intel helping pay for OEM product advertising on TV. If Intel wanted to advertise itself, they certainly could, but it’s way better to make your brand inseparable from the product it’s in, and it has worked pretty well for decades. If there are competing APIs, who do you think is going to get the most attention when the developer resources get assigned?
I know. This makes me rather disgusted by the current state of the gaming industry, to be fair.
Posted on Reply
#37
Veseleil
Oh my, more Optiscaler options incoming.
Posted on Reply
#38
Craptacular
Darmok N JaladIf you mean Glide, it was great, but poorly supported by developers. Quake II had Glide (and 3Dnow!), and performance skyrocketed, but no other games I played offered it. Then 3DFX hit hard times and sold their IP to NVIDIA, who has apparently never did anything with that tech.
I mean GeForce FX is based on 3DFX IP and its designers.

Half-life and Unreal had Glide API support, in fact there were over 190 games that supported glide api: www.pcgamingwiki.com/wiki/List_of_Glide_games
Posted on Reply
#39
Let3r
Darmok N JaladIf you mean Glide, it was great, but poorly supported by developers. Quake II had Glide (and 3Dnow!), and performance skyrocketed, but no other games I played offered it. Then 3DFX hit hard times and sold their IP to NVIDIA, who has apparently never did anything with that tech.
What exactly is Glide?
Posted on Reply
#40
ISSEnterprise
We are approaching to a possible future where the entire 'video games' are generated by AI models real-time.
Posted on Reply
#42
john_
kazuvikingThe upscaler part which is the most important haven't been upgraded in 2 years.
Still better than FSR ( I see Nvidia fans putting this post in their bookmarks). XeSS is considered good enough and to be fair, throwing resources there to improve visual quality by 5% and performance by another 5%, doesn't makes much sense when you are Intel and the Gaming GPUs aren't a big part of your business. Especially when you are concentrating on iGPUs and mobile handhelds where the current quality of XeSS is probably more than enough. On the other hand throwing resources on a feature that can double the framerate, makes total sense. Intel did the opposite here to Nvidia. They did the correct thing. Nvidia gave quality to older RTX cards, because it knew that quality without performance means upgrades. Intel gave performance to it's customers, showing them that their cards are also future proof and that's a reason for those owing an Intel (i)GPU to choose again an Intel (i)GPU in the future.
Posted on Reply
#43
kazuviking
john_Still better than FSR
Only better than fsr3 in the XMX mode.

XeSS have the worst performance out of all, even worse than transformer dlss. You need balanced upscaling to reach quality with fsr/dlss.
john_Intel gave performance to it's customers, showing them that their cards are also future proof and that's a reason for those owing an Intel (i)GPU to choose again an Intel (i)GPU in the future.
Barely any games have intels fg implemented and there wont be an intel dgpu for a very long time. Late 2028 is the earliest intels druid gpu can show up if it ever does.
Posted on Reply
#44
scottsoapbox
AusWolfNeural this, neural that... It's all basically just matrix operations. Don't they run on common instruction sets? Why do they have to be brand exclusive?
To. Sell. Their. GPUs.
Posted on Reply
#45
AleXXX666
TheinsanegamerNSeeing as how most major PC games are still more optimized for Nvidia cards and Nvidia tech, I think the results speak for themselves. The full version of far 4 diamond is in what, three games now?
if you use powerful enough red card and you DON'T need stupid DLSS thing then your gameplay experience will be perfectly same lmfao
The "optimization" is only "magic" "DLSS" option availability vs copy-paste crap like FSR and XeSS :D :rolleyes:
Posted on Reply
#46
Denver
TheinsanegamerNAMD could have kept it to themselves, but with their execution it would have withered and died. (Oh hi vega).

3DFX beat AMD to the low level API by 25 years.
That is completely and objectively false. GCN was very strong in compute, so much so that it was used as the foundation for CDNA.

AMD won numerous supercomputer contracts for this very reason. HBM would go on to become increasingly popular in server and compute hardware.

And a dead company doesn't beat anyone.

For those who enjoy gimmicks, AMD demonstrated an AI method to generate ray-traced lighting 5x faster.
gpuopen.com/learn/temporally-stable-generative-illumination/

Posted on Reply
#47
Prima.Vera
AusWolfNeural this, neural that... It's all basically just matrix operations. Don't they run on common instruction sets? Why do they have to be brand exclusive?

At this rate, developers will have to make two entirely separate versions of the same game, one for Nvidia, one for AMD. It's incredibly dumb, imo.
Indeed, that's why Microslop should had already provided a DirectX API for this. But since they are way over lazy and incompetent nowadays....
Darmok N JaladIf you mean Glide, it was great, but poorly supported by developers. Quake II had Glide (and 3Dnow!), and performance skyrocketed, but no other games I played offered it. Then 3DFX hit hard times and sold their IP to NVIDIA, who has apparently never did anything with that tech.
They took the SLI concept from 3DFx, multi-texturing, Tile-Based Deferred Rendering (TBDR), etc, while taking also more than 100 of 3Dfx's engineers....
Posted on Reply
#48
Luminescent
This news is basically some damage control, AMD has nothing to compete with DLSS5, they "leaked" to some known leaker the news so everyone can keep calm and still buy some AMD.
Just to remember, Nvidia launched DLSS in 2018, Amd launched FSR 4 in 2025, so we could be seeing something similar to DLSS5 on AMD in 5 to 7 years, yeah, that's why AMD is probably at 1% market share.
Posted on Reply
#49
Darmok N Jalad
Prima.VeraIndeed, that's why Microslop should had already provided a DirectX API for this. But since they are way over lazy and incompetent nowadays....


They took the SLI concept from 3DFx, multi-texturing, Tile-Based Deferred Rendering (TBDR), etc, while taking also more than 100 of 3Dfx's engineers....
I thought they grabbed TBR from the Kyro acquisition?
Posted on Reply
#50
chrcoluk
Why are people bothered about this feature?
Posted on Reply
Add your own comment
Nov 12th, 2026 06:33 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts