• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Microsoft Game Stack VRS update (Series X|S) - Doom Eternal, Gears 5 and UE5 - 33% boost to Nanite Performance - cut deferred lighting time in half

I never thought hardware Tier 2 VRS would be capable of this much. This is above and beyond what I expected. They're showcasing some pretty huge gains when it's implemented on a more advanced level throughout a game engine.
Let's see it in an actual game on a console first, so far when DF has pointed it out in games that currently use VRS on Series X they don't seem too happy with it.
 

elliot5

Member
Foveated rendering is a focus related solution. But for it to be implemented, the underlying engine requires the ability to render different levels of detail per frame.

In a 2d game you wouldn’t need to implement a Foveated technique, but you would exploit the different level of detail per frame ability in a way it makes sense for a flat screen.
It’s literally Variable Rate Shading. Adding in gaze tracking or just exploiting how the eye works in a 3D display (for fixed foveated) is what makes it work. It’s still VRS under the hood. Or Variable Rate Super Sampling to essentially do the inverse and hyper detail the focal point though that is more difficult
 

Shmunter

Member
It’s literally Variable Rate Shading. Adding in gaze tracking or just exploiting how the eye works in a 3D display (for fixed foveated) is what makes it work. It’s still VRS under the hood. Or Variable Rate Super Sampling to essentially do the inverse and hyper detail the focal point though that is more difficult
I don’t believe the ps5 hardware can alter the shading rate like the Xbox can - it’s static in ps5.

However both have the end goal in saving rendering time dynamically.

Xbox does it by adjusting the shading in real-time, whereas the ps5 culls the scene complexity of the area prior to any shading. Less complexity = less shading needed and savings are made. Indeed, ps5 reducing complexity upfront logically allows for deeper savings in rendering on the platform.
 
Last edited:

Fafalada

Fafracer forever
Is there really talk that foveated rendering for VR is applicable to regular TV games and is the superior technology to VRS? Not saying VRS is superior, but like they’re two different use cases of optimization and one requires an eye tracker lol
VRS isn't 'a' use-case - as a DirectX term, it refers to functionality that allows control over shading-quality independant from pixel-quality.
A use-case would, for instance, be using VRS to modify shading quality based on sensor-input from an eye-tracker.

And Foveated rendering doesn't describe any specific functionality or technology at all. It's an umbrella term that has been used to describe many different implementations/features and even use-cases. The common point is that to implement any form of Foveated rendering - you need to have non-uniform control over pixel(or shading) density that's decoupled from display resolution.
TLDR - whatever you use to implement Foveated rendering can indeed, be applied to flat-display use-cases.

Sorry for playing the terminology police - but technical discussion is nigh impossible if we just assign randomised meanings to terms.
 

elliot5

Member
I don’t believe the ps5 hardware can alter the shading rate like the Xbox can - it’s static in ps5.

However both have the end goal in saving rendering time dynamically.

Xbox does it by adjusting the shading in real-time, whereas the ps5 culls the scene complexity of the area prior to any shading. Less complexity = less shading needed and savings are made. Indeed, ps5 reducing complexity upfront logically allows for deeper savings in rendering on the platform.
Maybe not hardware accelerated via AMDs shading techniques for Tier 2 VRS but as demonstrated by Infinity Ward its still possible to some degree via software solutions. I’m not sure why you couldnt cull and VRS at the same time? It’s two different steps along the render pipeline
 

elliot5

Member
VRS isn't 'a' use-case - as a DirectX term, it refers to functionality that allows control over shading-quality independant from pixel-quality.
A use-case would, for instance, be using VRS to modify shading quality based on sensor-input from an eye-tracker.

And Foveated rendering doesn't describe any specific functionality or technology at all. It's an umbrella term that has been used to describe many different implementations/features and even use-cases. The common point is that to implement any form of Foveated rendering - you need to have non-uniform control over pixel(or shading) density that's decoupled from display resolution.
TLDR - whatever you use to implement Foveated rendering can indeed, be applied to flat-display use-cases.

Sorry for playing the terminology police - but technical discussion is nigh impossible if we just assign randomised meanings to terms.
Thank you, you’re correct. I was briefly trying to say it doesn’t make sense to claim FR is a superior method to VRS (pointing to the PSVR2 Unity gains) when that’s only really applicable to VR tech bc of the eye gaze tracking allowing for aggressive use of VRS (or some other method, but typically VRS from what I’ve read). If that makes sense.
 

Shmunter

Member
Maybe not hardware accelerated via AMDs shading techniques for Tier 2 VRS but as demonstrated by Infinity Ward its still possible to some degree via software solutions. I’m not sure why you couldnt cull and VRS at the same time? It’s two different steps along the render pipeline
Indeed, infinity ward claims in their gdc presentation their software solution surpasses that of the fixed function hardware as they gain more control over the result.

As far as tier 2 vrs co-existing with Sony’s bespoke geometry adjustment hardware, any additional benefit may be too diminished to bother with and not worth the die space.

Ms went pure off the shelf AMD, Sony decided to spend R&D effort and cost to build
something they believed to be off more benefit or they wouldn’t bother.
 
Last edited:
You do know that UE5 (matrix demo) uses prim/mesh shaders right? Yet they perform very similarly in UE5 at least. Why then do you argue that one is better than the other?
Only around ~5% of nanite geometries in ue5 demos thus far are making use of prim/mesh shader, the rest is done in compute shaders. So no nanite is not a good example of a mesh shader implementation.
S staticshock You realize PS5's Primitive Shaders aren't 1:1 the same as the Vega ones, right? AMD updated the Primitive Shader spec, and was never implemented in further Vega GPUs (to my knowledge).
This is simply not true, we have all the facts including patents from AMD. Its quite clear what Primitive Shader does.
What you are doing here is clearly making things up. Literally and creating this idea of a mytical supeior primitive shader that does amazing godly thing. Literally you are fabricating something from thin air.

Its funny that people laugh at others for pointing out the features of the Xbox, yet when they point out these features, they are pointing out DETAILED tech presentation with detailed analysis and numerical results from actual implementation.

You on the other hand, when you talk about PS5 features, its this mythical, secret fabricated thing with no actual fact or evidence.

Yet you laugh at others? WTF? How devoid of reality is that?
I'm not saying that to imply their Primitive Shaders are exactly like Mesh Shaders (there are obvious differences), but there's also nothing you can bring up refuting Matt Hargett's posts on the topic of VRS-like techniques on PS5, can you? ;)
We know they are not alike, we know there are differences and right down to the details what the differences are.
Lastly Matt didn't say anything about a VRS like technique. He's literally just saying primitive shader does what primitive shader was created to do.
Its not a VRS like technique.

Also chill on that Matrix demo already, that's got nothing to do with the topic 😂
It absolutely does. When you have a track record where everything you have said so far is incorrect and while there are 100% indisputable facts from even the creators of unreal engine and the demos, you still spread FUD and misinformation. It proves you are simply here for one reason.
 
Last edited:
Only around ~5% of nanite geometries in ue5 demos thus far are making use of prim/mesh shader, the rest is done in compute shaders. So no nanite is not a good example of a mesh shader implementation.

This is simply not true, we have all the facts including patents from AMD. Its quite clear what Primitive Shader does.
What you are doing here is clearly making things up. Literally and creating this idea of a mytical supeior primitive shader that does amazing godly thing. Literally you are fabricating something from thin air.

I never said Primitive Shaders were superior to Mesh Shaders 🤣. And I don't think their ability to control some of the level of detail for certain objects to be rendered later in the graphics pipeline in a way where such allows for more detail to be focused on other assets of a framebuffer, is "amazingly godly". It's just something in concept similar (regarding saving on needless rendering of certain elements) to what VRS aims to do, implemented at a different point of the pipeline.

That's all I've been saying really, and it makes sense because this is what PS5 Software Engineer Matt Hargett has basically said/alluded to in the past on Twitter.

Its funny that people laugh at others for pointing out the features of the Xbox, yet when they point out these features, they are pointing out DETAILED tech presentation with detailed analysis and numerical results from actual implementation.

You on the other hand, when you talk about PS5 features, its this mythical, secret fabricated thing with no actual fact or evidence.

There's this little thing called Road to PS5, with Mark Cerny spending nearly an hour talking nothing but technical features of the system in detail. You might've heard of it ;)

Yet you laugh at others? WTF? How devoid of reality is that?

We know they are not alike, we know there are differences and right down to the details what the differences are.
Lastly Matt didn't say anything about a VRS like technique. He's literally just saying primitive shader does what primitive shader was created to do.
Its not a VRS like technique.


Again, you're acting as though the Primitive Shaders in PS5 are exactly the same as the ones in the old Vega GPUs. They aren't. Sony's Primitive Shaders are based on an updated version which didn't make it into Vega GPUs and was disabled in RDNA1 GPUs, to my knowledge.

Also have not said that Primitive Shaders perform VRS, the whole time I've been saying that part of their functionality can serve as performing reduced processing-intensive rendering tasks similar in aim to what something like VRS does, at a different part of the graphics pipeline. I've never said what Primitive Shaders do in this regard are superior (or inferior, to that matter) to what VRS actually does, just that there is a similarity. And that's while also knowing the actual main usage of Primitive Shaders is quite different.
 

Three

Member
Is there really talk that foveated rendering for VR is applicable to regular TV games and is the superior technology to VRS? Not saying VRS is superior, but like they’re two different use cases of optimization and one requires an eye tracker lol
They both rely on the same underlying method of reducing shading rate on different parts of a render. Why wouldn't it apply to a 2D screen? VR is a 2D screen just rendered from 2 different viewpoints and put side by side.
 

Godfavor

Member

I don’t believe the ps5 hardware can alter the shading rate like the Xbox can - it’s static in ps5.

However both have the end goal in saving rendering time dynamically.

Xbox does it by adjusting the shading in real-time, whereas the ps5 culls the scene complexity of the area prior to any shading. Less complexity = less shading needed and savings are made. Indeed, ps5 reducing complexity upfront logically allows for deeper savings in rendering on the platform.
You can't save shading time from a texture, motion blur, particles or volumentrics using primitive shaders by culling the polygon because there is no polygon to shade. That's why VRS is not the same.
 
Last edited:

Panajev2001a

GAF's Pleasant Genius

Fafalada

Fafracer forever
Thank you, you’re correct. I was briefly trying to say it doesn’t make sense to claim FR is a superior method to VRS (pointing to the PSVR2 Unity gains) when that’s only really applicable to VR tech bc of the eye gaze tracking allowing for aggressive use of VRS (or some other method, but typically VRS from what I’ve read). If that makes sense.
So specifically to PSVR2/Unity FR example - it is an interesting question how that was implemented. PS5 does for instance, carry feature that PS4Pro introduced specifically to accelerate this type of VR workloads. But we don't know if that's what Unity demos used at all. You're right though - that even if using VRS specifically, comparing gains in that demo to gains in a flat-screen game with no eye-tracking doesn't mean anything.

That said - my point earlier, was that FR isn't a method but a collection of solutions/methods. Ie. VRS is just one of the possible solutions for FR, so at most - you can compare them by saying one is a functional subset of the other. There's no 'better or worse or equal' in that comparison.

As for 'what's the optimal solution' for FR - that's a longer discussion that depends on factors that can be highly variable from project-to-project. From general perspective you can say VRS is not it because it's only solving one part of the problem (shading resolution), so you'd always need something else in addition to VRS to solve FR adequately. Then there are other approaches that offer a more complete solution - but they come with different tradeoffs. But this is also a bit of an academic debate - optimizing for games isn't a 'game' of "what's your favorite tool" - you pretty much throw everything you can find (and can afford) at the problem, and if that means combining 1001 different optimizations, so be it.

TLDR - it's basically impossible to have a real conversation on 'what's optimal' without defining context (ie. what is the game/technology stack/asset workloads/device we're optimizing for). Making those claims in abstract without such context is pretty much just blowing hot-air, it doesn't matter what tech is being championed.
 
I.e no one should feel insecure about how their preferred console handles the issue of reducing needless details to optimize runtime performance, because regardless in what ways/quirks the systems implement them at the hardware level, they're relatively close enough in practice where only people hunting for visual evidence (like DF) will spot those differences anyway. The vast majority of gamers won't be able to blindly tell.

Plus it's not like devs can't run their own software-driven solutions, which they've been doing anyway, and both systems have more than enough hardware grunt to handle the overhead of a software solution*

*that isn't going to stop mid-gen refresh rumors or speculation though...although I suppose those are coming down to lack of hardware headroom for 4K 60 with full RT on top. Still don't know how that would be a great decision for mid-gen refreshes if it ends up minimizing the impact of 10th-gen consoles.
 
VRS can just be "turned on" but VRS gives significantly better results when it's plumbed throughout the entire game engine. It can do more than just the more basic use case for it if implemented throughout the game engine. Coalition already said this last year.


What we are seeing right here is an example of some of the very things Coalition was mentioning could be possible with a deeper integration of VRS into a game engine. That Variable Rate Compute Shader stuff is mighty impressive, for example.
So are you talking about running VRS continuously throughout the whole game?
VRS isn't a tech that was introduced to give sustained performance gains over a full game. It's like DRS for Shading.
It's only used in stress areas of the game where performance is dropping. Once that area is cleared VRS is turned off.
Why would devs import high quality textures into a games memory, and then have it drawn by the GPU to then only reduce the quality of those assets in game. They would just import lower quality textures and save on the RAM and GPU usage in the first place and not need VRS.
 

Riky

$MSFT
So are you talking about running VRS continuously throughout the whole game?
VRS isn't a tech that was introduced to give sustained performance gains over a full game. It's like DRS for Shading.
It's only used in stress areas of the game where performance is dropping. Once that area is cleared VRS is turned off.
Why would devs import high quality textures into a games memory, and then have it drawn by the GPU to then only reduce the quality of those assets in game. They would just import lower quality textures and save on the RAM and GPU usage in the first place and not need VRS.

The opening post shows how it's used throughout the game, the point of that is that you need less reliance on DRS, the Doom Eternal Devs talked about this at length.
 
Last edited:
The opening post shows how it's used throughout the game, the point of that is that you need less reliance on DRS, the Doom Eternal Devs talked about this at length.
So they are relying on gaining a higher resolution to "hide" the image degradation from VRS?
I guess we will see the real world results once more games are released using it.
Up until now DF has been able to pick out the use of VRS in games, so let's see moving forward.
It's a great tool to have the use of and it's going to be a win for Xbox this generation, especially as games become more demanding.
 

Riky

$MSFT


Redtechgaming video on VRS and the presentations. Nothing really new but he does simplify parts for people to understand.
 
Top Bottom