Support NeoGAF

[yamaci17]

FSR and FSR 2.0 use the same LOD system. All these upscalers, require adjustment to LODs.
The difference is the use of temporal information by FSR 2, DLSS 2, XeSS, TAAU and TSR,

not necessarily, fsr 1.0 did not use the LOD system at all. it directly drops EVERYTHING to the rendered resolution and applies a sharpening filter

FSR 1.0 can be literally used as a post process filter by RPCS3, Xenia, emulators, many more software etc. Engine level FSR is no different than post process FSR. It simply does nothing with engine itself. Its just a simple way of reducing the resolution directly as if you're choosing a lower resolution and apply a sharpening filter.

you can go find any fsr 1 and fsr 2 game (there are some) do comparisons between lods both respectively at quality settings and you will quickly realize this is true

 

[winjer]

not necessarily, fsr 1.0 did not use the LOD system at all. it directly drops EVERYTHING to the rendered resolution and applies a sharpening filter

FSR 1.0 can be literally used as a post process filter by RPCS3, Xenia, emulators, many more software etc. Engine level FSR is no different than post process FSR. It simply does nothing with engine itself. Its just a simple way of reducing the resolution directly as if you're choosing a lower resolution and apply a sharpening filter.

you can go find any fsr 1 and fsr 2 game (there are some) do comparisons between lods both respectively at quality settings and you will quickly realize this is true

None of these upscalers affect the LOD system, it has to be adjusted accordingly. Only a bad implementation of an upscaler, will have the same LODs as default.
Emulators can also do some of this, like changing MipMaps LOD. But not every effect is accessible to the user.

 

[yamaci17]

None of these upscalers affect the LOD system, it has to be adjusted accordingly. Only a bad implementation of an upscaler, will have the same LODs as default.
Emulators can also do some of this, like changing MipMaps LOD. But not every effect is accessible to the user.

they do affect the lod system.

forza 5 4k dlss perf

Imgsli

game literally uses better textures, higher level of lods at 4k/dlss performance. both are supposed to be 1080p. fsr 2 is similar story but I did not record. in game res. scale uses something like FSR 1. when you use in game res. scale to get back to 1080p, it results exactly like how a 1080p native image look

Imgsli

in this example you can literally see 4k/dlss performance uses better lods, higher quality assets, and even draw distance is affected by "master" upscaled resolution.

 

[winjer]

they do affect the lod system.

forza 5 4k dlss perf

Imgsli

game literally uses better textures, higher level of lods at 4k/dlss performance. both are supposed to be 1080p. fsr 2 is similar story but I did not record. in game res. scale uses something like FSR 1. when you use in game res. scale to get back to 1080p, it results exactly like how a 1080p native image look

Imgsli

in this example you can literally see 4k/dlss performance uses better lods, higher quality assets, and even draw distance is affected by "master" upscaled resolution.

MipMaps are dependent on native screen resolution. This defines the distance at which a mipmap level is changed.
So at lower screen resolutions, the next level of mipmap will be loaded at a shorter distance from the camera.
When we upscale an image, be it with a spatial or temporal upscaler, the game only knows about the base resolution.
So a game rendering at 1440p using an upscaler, at 66% screen resolution, will use the Mipmap values of 960p.
This will look bad, because the game will load the next level of mipmaps too soon, resulting in blurrier textures, closer to the camera.
So the developer has to manually tune the mipmap bias, to force the game to load the mipmap bias adequate to the output resolution.

You might have seen Alex complaining about some games not properly adjusting the mipmap bias. And then talking about using the gpu control panel, to offset the mipmap bias.
This is something that should be done by the devs, not the users.

Now game engines are also using screen resolution to scale other parts of the rendering pipeline.
For example, nanite renders triangles in relation to pixel size.

Here is a page from the DLSS manual, on implementing mipmap ias.

 

[twilo99]

Is there a video showing the improvements on the PS5? I’ve only seen Xbox so far

 

[YeulEmeralda]

Wide open field vs an actual fucking metropoliton.

Crowd density would literally melt the Switch.


I'm gonna be honest here I thought it was impossible to make a somewhat realistic open world city but CP2077 did it.

 

[bbeach123]

they do affect the lod system.

forza 5 4k dlss perf

Imgsli

game literally uses better textures, higher level of lods at 4k/dlss performance. both are supposed to be 1080p. fsr 2 is similar story but I did not record. in game res. scale uses something like FSR 1. when you use in game res. scale to get back to 1080p, it results exactly like how a 1080p native image look

Imgsli

in this example you can literally see 4k/dlss performance uses better lods, higher quality assets, and even draw distance is affected by "master" upscaled resolution.

This is TAA doing god work for ya (sarcasm) , it ruined textures , detail , foliage , tree , grass n stuff .

DLSS just undo what TAA did .

 

[Arioco]

Is there a video showing the improvements on the PS5? I’ve only seen Xbox so far

I've posted one on this very thread, running on PS5 and in HDR.

 

[In Cold Blood]

So where are we at with CP 2077 on XSX?
Is now the time to dust it off and play it, or do they still need more patches?

 

[Tomeru]

So where are we at with CP 2077 on XSX?
Is now the time to dust it off and play it, or do they still need more patches?

Its fucken great, dust it off already!

 

[Crispy Gamer]

Is there a video showing the improvements on the PS5? I’ve only seen Xbox so far

Digital Foundry video is up

 

[Modi1984]