Hi, I saw that URP TAA was added to the public graphics repository few days ago.
But seems that TAA does not appear in 2023.1.a12.
Will TAA be available in 2023.1?
Hi, I saw that URP TAA was added to the public graphics repository few days ago.
But seems that TAA does not appear in 2023.1.a12.
Will TAA be available in 2023.1?
Hey there again!
It was actually added to In Development on the roadmap a while back. But yes, we are working to land it in 23.1.
Will it be based on/similar to HDRP’s TAA?
Based on experience, HDRP’s TAA has been in a bad shape until recently with severe ghosting fixes, the addition of a jitter parameter, and most recently great new sharpening quality options (contrast adaptive sharpening & post sharpen).
URP TAA is not directly based on HDRP TAA, but they do share some tricks and are quite similar.
URP TAA is mobile oriented and simplified. The interface is quite minimal and many settings are baked into the presets.
Post sharpening is available by using FSR. Which you can run at 1:1 scale for the sharpening.
I hope it’ll at least have a jitter parameter exposed.
Not sure why URP’s TAA is mobile oriented at the expense of everyone using URP for PC and consoles, most mobile video games will actively stay away from TAA. I’d go as far to say TAA is expected for PC/consoles, but not mobile.
Hopefully this doesn’t mean it’ll be any worse than built-in TAA, I thought the long wait for TAA would mean a good implementation from the start.
Currently most popular URP games are using assets to make up for bad URP features, because they put mobile before PC, for example URP SSAO gives worse visuals and slower performance compared to HBAO, and compared to HDRP’s SSAO.
And so, many popular URP games end up using HBAO (for example: BONELAB, Lost In Random, Craftopia, etc)
We won’t always have an alternative in the asset store, especially an updated one.
Please don’t hardcode settings, locking things like this will literally always be bad for users. For what. Simplicity?
Considering all URP/SRP settings I think TAA parameters are not something to worry about.
If you’re worried about complexity, put them under advanced settings. Maybe an alert not to change any of them unless you know what you’re doing.
HDRP’s TAA instantly became much more usable the moment they implemented a jitter parameter.
Even built-in exposed a few parameters, as you’d expect from any TAA implementation.
I presume it’s cause the mentality they expect people to use is “URP and Hdrp in the same project with” coexistence", chose one at build time per platform.
Except in the thread about said coexistence it was hinted we’ll still need two different light rigs for each scene because the light intensity and falloff function completely different between the two SRPs and they didn’t thing of unifying that.
I think it was said that that’s currently the case. Cause they asked how to deal with it right now.
https://discussions.unity.com/t/895665
In this question they said while there is some concern if users will understand it they will be unified or at the very least similarly physically based.
As such the issue of converting values would be mitigated.
This, the PPI is so insane on mobile (less so on tablet), is TAA really a thing for it?
I also don’t understand why delay implementing TAA to make it work on for mobile, where MSAA is cheap due to TBR GPUs and full screen passes are expensive due to the high resolutions and lower memory bandwidth.
It’s the same deal as SSAO, another effect that tanks mobile performance and ends up being compromised on capable platforms due to having to be implemented with mobile limitations in mind (aka: using fragment shaders instead of vastly faster compute shader implementations).
@ali_mohebali I have published 5 mobile games and never used TAA for mobile games. I even didnt see any team who use TAA for mobile.
URP’s focus is mobile and untethered first. If workflows/features we provide can scale and be supported on untethered devices as well, we make sure the techniques used will be able to scale across.
Out of curiosity, what is the reason you wouldn’t utilise HDRP instead of URP if you are only targeting PC and Consoles? Or are you saying you need URP because you intend to target both the mobile and PC/Console market?
There are customers who target only a higher-end spectrum of untethered devices for their content. If you don’t need it for your mobile game, there are other options we provide for AA in URP. This is why we left TAA to be part of the last Built-in functionality parity items we add to URP. It has been lower in priority compared to other gaps and features we wanted to support (and its also based on requests and feedback we have received).
What? I thought URP was a scale-able solution from mobile to PC, not mobile focused?
HDRP has numerous downsides from high requirements to a very real-world focused lighting & camera pipeline. It is not for non-realistic/stylized games as far I can see (would need heavy post effects).
Various game and none-game projects require a scalable pipeline to support a range of devices & PC level - that is what we want. So, URP from mobile to high-end PC.
Let me quote the manual for you @ali_mohebali
“The Universal Render Pipeline (URP) is a prebuilt Scriptable Render Pipeline, made by Unity. URP provides artist-friendly workflows that let you quickly and easily create optimized graphics across a range of platforms, from mobile to high-end consoles and PCs.”
URP can more easily scale down, particularly in the CPU department. Although HDRP performance has been seeing many improvements. Also, URP has a different visual look to HDRP which sometimes is a better fit for certain projects. Although this might vary from one person to the other.
Personally, I’m mainly using HDRP from the start for my main project but I was hoping URP would be a viable alternative for future projects or project switches for whatever reason.
The Issue I have is how URP imo isn’t aiming to be like built-in at all, it’s a watered down version of built-in with as you mentioned complete focus on mobile, everything else comes after. If this was the goal, why not keep it as the lightweight render pipeline? LWRP has a description that better aligns with this goal.
I didn’t know the mindset from URP developers was you shouldn’t use URP unless you’re aiming for mobile or switch release.
And I understand the high-end part, still begs the questions, why simplify it and hardcode settings?
TAA is very game specific, most people will need to play with the settings to get the right experience for their particular project. Some might want to reduce TAA effect, just to do near enough for geo/spec aliasing.
I’d also like to entertain the idea of quality modes, one for mobile, the other for PC/consoles.
Same with SSAO. Although I understand this comes at a cost, but it feels like the most universal solution to me.
Doesn’t sound good - feel like staying on built-in because URP and HDRP are awkward enough add-ons already and if they lack the scalability and features then not good…
Is it really Universal if a huge number of techniques are never going to be implemented into it because some lowest common denominator can’t support it? That’s really really painful to hear as someone working on a PC game with URP. I want to support computers that aren’t high end, so I have to avoid HDRP, but now in URP I’m a second class citizen…
I know this was a marketing decision not a engineering decision, but I think it’s clear that URP should have stayed as LWRP.
I haven’t used HDRP but I’d be surprised if it is impossible to work on a low end PC.
That comes down to more than available effects.
For me the only reason I chose URP was platform versatility cause I want to target mobile. But if “coexistence” solves that I will almost certainly switch to HDRP for PC.
But there’s also games like Lego builders tale that allows switching between URP and HDRP for low end.
So that could be an option too?
Imo tho TAA is now obselete tech in light of DLSS, FSR 2.0, and XeSS.
None of those are antialiasing techniques, they are upscaling techniques and they all designed to work best when fed antialiased source material.
No they all replace TAA as most (actually probably all) games don’t allow TAA while these features are on.
This is why stuff like XeSS has an “Ultra Quality” because the effect create a nicer anti aliasing without the usual downsides of TAA plus performance gains.
In this video below XeSS compares itself as a replacement for TAA, not simply an upscaler.
https://www.youtube.com/watch?v=frlXry38tJo
These are not just upscaling techniques, they are also AA solutions that far exceed the appearance and performance of TAA, FXAA, and sometimes even other solutions.
Even if you can play a game at 60fps no problem, adding XeSS or other solutions with “quality” or “Ultra Quality” will regularly provide a higher quality image than even native resolution especially compared to TAA.
Further evidence is this comparison comparing DLSS, to FSR 2.0, to Native with TAA.
https://www.youtube.com/watch?v=y2RR2770H8E
Granted it’s still not perfect for 100% of scenarios, but it’s still mostly an improvement on TAA and will only be more so with time.
TAA will still be around for a long time, not everyone has an RTX card.
Intel’s share of the dedicated GPU market is around zero, it’ll take a while for it to be worth mentioning.
FSR 2.0 is not in unity, and requires DX12. Also, FSR 2.0 uses TAA as usual, just a good implementation of TAA.
You need a general anti-aliasing technique for all your users, the ones who have an RTX card can choose DLSS/DLAA.
Most GPUs in the wild are still GTX, although RTX is a close second.