Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

[trinibwoy]

no confirmation by devs just obvious deduction (df thinks same)

Well in the wccftech interview the devs said the speed up is primarily due to the new console platforms / SSDs so that can be considered confirmation.

 

[snc]

Well in the wccftech interview the devs said the speed up is primarily due to the new console platforms / SSDs so that can be considered confirmation.

Ok, so thats confirmstion, lmao

 

[chris1515]

Ok, so thats confirmstion, lmao

They didn't use any I/O on Xbox or PS5 or they don't know how to use it. This is not better than the game running on PS4/PS4 PRo with a SSD. Some people can load games much faster than 4a games.

 

[snc]

They didn't use any I/O on Xbox or PS5 or they don't know how to use it. This is not better than the game running on PS4/PS4 PRo with a SSD. Some people can load ames much faster than 4a games.

That was joke :d

 

[chris1515]

That was joke :d

I know I just answer because the idea than a games load in 40 seconds and the games is fully using the PS5 I/O is just stupid or 4A games are bad programmer because other games using raytracing can load around 1 second.

 

[snc]

I know I just answer because the idea than a games load in 40 seconds and the games is fully using the PS5 I/O is just stupid or 4A games are bad programmer because other games using raytracing can load around 1 second.

just remember its free patch so still kudos for 4a games

 

[chris1515]

just remember its free patch so still kudos for 4a games

This is not against 4A games, this is great this free upgrade but against the idea of a games loading in 40 seconds using PS5 or Xbox Series I/O. When the loading time are so long, it is easy to understand they don't use new API.

 

[Globalisateur]

there is only so much I can say - but as far as I know (and I have it on reliable source) the reason why Xbox Series X and S are getting drops in the open world (like the ones I highlight in the Xbox video very poignantly), is not a GPU load issue to due proper utilisation (aka GPU is overtasked due to what is happening rendering wise), but something else that is in all likelihood going to be patched. But in a place like Yamantau where it is definitely GPU load and proper utilisation, where in that one scene you see Xbox Series X dipping to 1134p and hitting around 57-58 fps while PS5 is 1015p at 60 fps, XSX could use a small res drop (to 1080p probably) to clear up that small instability for sure. But then again, with VRR that is completely invisible... so...

Thanks for that very precise data. So that would make about 1080p vs 1015p. You know what that means? It means if XSX had the same stable framerate (with a more aggressive DRS), then the resolution gap would be reduced to about 12%.

Why didn't you put that in the article? That's a very interesting comparison in one of the most GPU demanding scene (without any I/O shenanigans). It's a great benchmark for both systems.

 

[iroboto]

Thanks for that very precise data. So that would make about 1080p vs 1015p. You know what that means? It means if XSX had the same stable framerate (with a more aggressive DRS), then the resolution gap would be reduced to about 12%.

Why didn't you put that in the article? That's a very interesting comparison in one of the most GPU demanding scene (without any I/O shenanigans). It's a great benchmark for both systems.

That scene only represents the absolute worst case scenario for the game. Not the average or best. Not is it representative of the majority of time on the game. But more importantly they didn’t measure XSX dropping down to 1080p. There is nothing they can do about frame rate issues.

Insisting that this should be 1080p would be false reporting in this case since that is a value they did not collect.

 

[Globalisateur]

That scene only represents the absolute worst case scenario for the game. Not the average or best. Not is it representative of the majority of time on the game. But more importantly they didn’t measure XSX dropping down to 1080p. There is nothing they can do about frame rate issues.

Insisting that this should be 1080p would be false reporting in this case since that is a value they did not collect.

Using one of the worst frame is how they did in Hitman 3 (and that was a cutscene, so even less representative of gameplay), both games had different resolutions so they took that difference into account (by measuring number of pixels rendered) and they wrote a whole article around it.

I'd say that scene in Metro is a more representative (more fair) benchmark because it's actually gameplay. And don't forget that in order to have locked framerate your usually need to have some margin above 60fps.

 

[iroboto]

Using one of the worst frame is how they did in Hitman 3 (and that was a cutscene, so even less representative of gameplay), both games had different resolutions so they took that difference into account (by measuring number of pixels rendered) and they wrote a whole article around it.

I'd say that scene in Metro is a more representative (more fair) benchmark because it's actually gameplay. And don't forget that in order to have locked framerate your usually need to have some margin above 60fps.

They had no choice though with hitman. There were no dips period except for that singular measurable point. @Dictator already explains it’s pointless to try to measure up the consoles against PC because the settings are entirely different.

the point of comparison was to get an idea of where the consoles stood versus PC. And more importantly or teach the audience what the possible causes of those dips could be. That can’t be done with this case.

and once again he did not inject unreported values into the Hitman video.

 

[trinibwoy]

[Wccftech] Did you evaluate the implementation of other DirectX 12 features like Mesh Shading and Sampler Feedback when developing Metro Exodus Enhanced Edition? Are you planning to use them in future titles?

Ben Archard: As far as Mesh Shaders, they're not really something we had any need to use because they essentially change the way that you run your geometry pipeline and we already have a very mature geometry pipeline. Maybe we can experiment with them, we're not ruling them out, but there weren't strictly speaking necessary for this project.

source:
https://wccftech.com/metro-exodus-e...are-much-more-powerful-than-we-first-thought/

It makes sense that they're not too excited about mesh shaders since it would complicate their RT pipeline. They sound very confident in their existing geometry pipeline too but I don't know if they should be. Granted it is a last generation game but the assets and environment in Exodus weren't sufficiently impressive to be dismissive of the benefits of mesh shaders.

 

[Proelite]

Thanks for that very precise data. So that would make about 1080p vs 1015p. You know what that means? It means if XSX had the same stable framerate (with a more aggressive DRS), then the resolution gap would be reduced to about 12%.

Why didn't you put that in the article? That's a very interesting comparison in one of the most GPU demanding scene (without any I/O shenanigans). It's a great benchmark for both systems.

Lowest and highest boundaries are not the best measure. It could be in one scene the X struggles due to geometry / ROP bottlenecks, where in other scene the PS5 struggles due to compute / bw bottlenecks like here.

I made an old apple to apple comparison from the trailers release a few weeks.

PS5:

Xsx:

PS5 is 1080p while Xsx is 1224p.

 

[Kugai Calo]

It makes sense that they're not too excited about mesh shaders since it would complicate their RT pipeline. They sound very confident in their existing geometry pipeline too but I don't know if they should be. Granted it is a last generation game but the assets and environment in Exodus weren't sufficiently impressive to be dismissive of the benefits of mesh shaders.

RT pipeline is pretty much separated from rasterization. In rasterization after the geometry stage much of the world space information is thrown away, whereas RT needs to keep them. They probably just don't need the performance and flexibility of mesh shader yet.

 

[trinibwoy]

RT pipeline is pretty much separated from rasterization. In rasterization after the geometry stage much of the world space information is thrown away, whereas RT needs to keep them. They probably just don't need the performance and flexibility of mesh shader yet.

The pipelines are independent yes but not the art assets. You would need to ensure that whatever LOD scheme you come up with is compatible with both mesh shaders and RT.

 

[Proelite]

yeah on neighter console, it seems they used what is availabe in the api
"...There's a lot of that is handled natively, or pretty much all of that is handled natively on the platform API side. And the fact is that those SSDs are just super fast..."

intresting bit here

"
[Wccftech] Did you evaluate the implementation of other DirectX 12 features like Mesh Shading and Sampler Feedback when developing Metro Exodus Enhanced Edition? Are you planning to use them in future titles?

Ben Archard: As far as Mesh Shaders, they're not really something we had any need to use because they essentially change the way that you run your geometry pipeline and we already have a very mature geometry pipeline. Maybe we can experiment with them, we're not ruling them out, but there weren't strictly speaking necessary for this project.

Sampler Feedback, again, it's something we'll look into. I like the look of it, but it needs to be experimented on. In Metro Exodus Enhanced Edition we do have Variable Rate Shading, a Tier 1 implementation that we use to reduce the actual number of pixels drawn for our transparent effects. That helps performance a little bit, but we haven't gone fully into Tier 2 VRS yet.

Interesting.
No wonder the transparency stuff looks pixelated on the S. Tier 1 is extremely limited compared to tier 2.

I don't think sfs wasn't even available as part of the SDK during the time they were making this upgrade.

 

[Dictator]

I know I just answer because the idea than a games load in 40 seconds and the games is fully using the PS5 I/O is just stupid or 4A games are bad programmer because other games using raytracing can load around 1 second.

I hope people do not think that just because loading times. I think we here all know it is a matter of project priorities and release times as to what gets technical considering and what not. This project like their Blog said, was all about getting and making sure RtGi was performant on consoles so that they can prove it works and use it for future projects not just on PC. Gosh how terrifying it must have been to be willing to make that bet and not being sure what RT performance would be there at all!

 

[Clukos]

This is not against 4A games, this is great this free upgrade but against the idea of a games loading in 40 seconds using PS5 or Xbox Series I/O. When the loading time are so long, it is easy to understand they don't use new API.

Could be they ran out of time, recompressing the data and using new APIs needs extra dev / qa time. Probably a bit late now but for a free patch this is very impressive.

 

[chris1515]

I hope people do not think that just because loading times. I think we here all know it is a matter of project priorities and release times as to what gets technical considering and what not. This project like their Blog said, was all about getting and making sure RtGi was performant on consoles so that they can prove it works and use it for future projects not just on PC. Gosh how terrifying it must have been to be willing to make that bet and not being sure what RT performance would be there at all!

This is a free upgrade. I understand. It is not against 4A games. It is against people thinking a game loading in 40 seconds use DirectStorage on Xbox Series X and PS5/I/O. I don't understand how someone can think this.

I am sure when you talked with 4A games you did not ask if they use the I/O API of current gen consoles. You understood with the loading times.

 

[Globalisateur]

Lowest and highest boundaries are not the best measure. It could be in one scene the X struggles due to geometry / ROP bottlenecks, where in other scene the PS5 struggles due to compute / bw bottlenecks like here.

I made an old apple to apple comparison from the trailers release a few weeks.

PS5:

Xsx:

PS5 is 1080p while Xsx is 1224p.

Nice work, but that's from an old trailer, not current state of game.