Seems like it takes forever
is dlss 1.0 available?
3 Likes
What I’ve seen dlss is not available. I saw a video where it was in Unreal Engine, but haven’t seen it in Unity.
I think for Unity to skip on DLSS 2.0 would be a big mistake.
- It gets rid of TAA’s ghosting and texture blurring issues
- It brings 4K gaming to a realm of possibility.
But, given that PS5 and Xbox SX are AMD based, there is a big chance that Unity will wait for AMD’s DLSS which probably will be hardware agnostic. This probably will make Unity’s rendering far behind Unreal once again (Unreal already has DLSS) until AMD’s counterpart is released. (I expect 1 year for AMD to release it, another year for Unity to implement)…
Let’s just hope that my predictions are wrong and Unity implements DLSS 2.0 right away.
2 Likes
I was just wondering if the HDRP team is looking into DLSS 2.0.
It looks like DLSS 2.0 is superior to TAA in every possible way.
- DLSS2.0 gets rid of ghosting
- 4K gaming performance
- transparent material rendering is stabilized.
Here is a video for the interested.
4 Likes
As far as I understand, DLSS 2.0 no longer needs training and is a general-purpose tech. DLSS prior to 2.0 (so, 1.9 and earlier) used to be like you said as it needed training on a massive scale which was not feasible for us.
Here is nVidia’s main website with more info:
and here is the exact exert:
One Network For All Games - The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalized network that works across games. This means faster game integrations, and ultimately more DLSS games.
1 Like
And I’ll merge these threads. The order of merged posts are chronological.
1 Like
Yeah, does it work on AMD? I would doubt it. But ms and sony are AMD.
I’d like to suggest to Unity that they have a backend like XR, where each manufacturer is responsible for specific implementations. So far Oculus, Valve etc are all maintaining their own XR implementations in Unity. Unity just provides an abstracted interface for many possible features.
If applied to SRP, that would give us DLSS 2.0. It’s also why UE4 can offer it, source and all.
1 Like
Only on Nvidia RTX gpu’s, so even majority of Nvidia gpu’s don’t support it. It needs the tensor cores hardware to work which only exist on RTX lineup.
I looked at the UE4 program and you need to apply to access it fully. Basically this means they hand out access to it per project basis atm.
It’s what I expected. So you’d probably need a branch of an SRP to make it work, and this would likely just be HDRP. And you’d need to be applying for access. I expect this restriction will lift though.
So is raytracing, but we are moving forward with the expectation that it will be widely implemented. Although I am not too sure how AMD’s tensor core counterpart will come out. I just hope Unity is ready to implement AI-enhanced upscaling as soon as it is available.
Well, there’s a raytracing support on some non RTX gpu’s from nvidia, altho they are too slow to be used still. Also AMD’s next gen cards will have DXR support (but I doubt we’ll see dlss compatible tensor cores from amd as afaik it’s a proprietary nvidia implementation).
AMD’s simply waited a couple of years not chasing nvidia so that they might use general purpose compute with a few tweaks to achieve DXR support. Just a matter of having fast enough memory and enough of it for it to make sense.
Unique hardware in GPUs can be a burden or a blessing depending on the time, I guess. In any case I tend to prefer gpu compute tech because even mobile can nibble at it 
while dlss is nVidia’s AI enhanced upsampling isn’t. My guess is Microsoft would probably include it in their DX12 soon, or at least I hope. Given that both AMD and nVidia are trying to achieve 4K, it seems inevitable as the performance/quality is quite outstanding. The hardware part, tensor cores as you mentioned worries me a bit though
This is a GAME CHANGING FEATURE!.. Unity when will you have support for DLSS 2.0? Also, don’t forget about proper Raytracing support?
AMD has had a solution available for almost an entire year known as Radeon Image Sharpening.
https://www.techspot.com/article/1873-radeon-image-sharpening-vs-nvidia-dlss/
Ryiah, the two are vastly different. While AMD wants people to see it as their counter part solution, CAS is just upscaling done at its best at most. A dynamic pixel filter to be more exact.
DLSS is a reconstruction of image through deep learning accelerated by RTX hardware. There are obvious performance benefits but there are huge image benefits when it comes to transparency and other related images.
The two are different as finite state machines are to AI neural networks. They are not comparable in anyway other than the fact that AMD wants it to look like it.
Also, as many of you have noticed DLSS is improving over time as the nvidia training data is getting added with servers, whilst CAS is just there not improving. Maybe a bit of an improvement but not like DLSS. It is literally learning to be better every second. A little scary to be honest.
People laughed at DLSS 1, and rightfully so, then DLSS 2 just surprised everyone and with the upcoming DLSS 3, it is going to replace TAA as nVidia plans it to work for all games with TAA. My guess is that anyone with an RTX card will choose DLSS over TAA as it will be widely implemented (DLSS3), faster than TAA and better looking (no ghosting) and also superior texture sharpness.
IMHO, the big question at the moment is when AMD will finally respond to nVidia’s AI related stuff. It is definitely a losing game over time as nVidia’s implementation improves every second like earning interest. When AMD finally implement their own DLSS, it will take as much and probably more time than nVidia. So, yes, DLSS is very significant, and will increase its influence very quickly. AI enhanced imaging is the future and this won’t change regardless. I just hope AMD will come out with their solution ASAP.
2 Likes
DLSS is now available to use in VR as well.
This article explicitly calls out unity as missing any sort for DLSS, support in VR would be a massive help given the extra processing power needed for VR titles.
I’ve avoided this thread for a while, but I want to add a bit side commentary to this whole discussion.
Nvidia’s DLSS 2.0 is amazing, and very different from AMD’s Radeon Image Sharpening. The later is more comparable to Nvidia’s Image Sharpening feature that’s been available for years, though RIS is superior to Nvidia’s implementation. However most games that use TAA already have something like AMD’s Radeon Image Sharpening implemented to counter the blurring artifacts of TAA, including Unity.
People here seem to be under the impression that it’s Unity’s decision to not include DLSS. DLSS 1.0 required per-game training, so it was never going to be an engine only feature. DLSS 2.0 was added to Unreal, but originally only as a non-public branch that you have to ask Nvidia permission to access. Nvidia’s clearly been working closely with Epic on this for a while, and it’s the only publicly available engine it’s been implemented in. It’s far more likely that Nvidia themselves have chosen to limit themselves to working with Epic on this for the early public version of DLSS. Plus Unreal is mainly a deferred renderer with TAA enabled by default. The vast majority of users will be setup already to make use of DLSS.
In contrast Unity is still primarily a forward renderer, with deferred being optional, and TAA being yet another option that people may or may not use. It makes less sense for Nvidia to spend effort supporting Unity when most titles using it won’t be in a position to take advantage of the tech.
However, I do expect to see Unity’s HDRP get support for DLSS, probably some time after DLSS 3.0 has matured.
Note: I have no insider info on this. I am not a Unity employee, nor have I asked anyone at Unity this (not that they’d be able to say anything). It just doesn’t seem fair to toss Unity under the bus here without any kind of information either way. On the other hand they do tend to shy away from vendor specific anything.
8 Likes
The current process to get DLSS is still tricky. You have to first get the SDK which exclusive to a selected number of unreal engine users ou custom engine users. If you are selected, you can use the SDK but you will get a watermark. Once your game is ready, you can ask nvidia to remove the watermark.
I think nvidia want to control/limit the usage of DLSS because it’s still early tech and they don’t want the tech to have a bad reputation. It’s super tricky to access this tech unless you are a well known developer.
I guess it won’t be a widely accessible SDK until they release DLSS 3.0 or something. For sure Unity has discussed this tech, it’s just not ready to be used by wide range of users.
AI upscaling is the future, Microsoft + AMD are working on it for consoles, the tech will become widely available, probably not until 1 or 2 years still.
4 Likes