NIS - when NVIDIA fixed what FDev wont.

[Morbad]

I remember a lot of noise when the DLSS and FSR were announced. But I don't remember anything like that about the CAS.

CAS came after the original DLSS and is a pretty basic post process spatial upscaler + sharpen filter that wasn't much different from filters already in existence. It was definitely in the media, and often compared favorably against DLSS 1.0, but it certainly wasn't marketed as well, and was clearly inferior to most DLSS 2.0 implementations.

FSR has been pushed harder and has attracted more attention as it's being presented as a more accessible DLSS near-equivalent.

It seems to me that the inclusion of different scaling coefficients in CAS is only in the Elite.

There are other games using FidelityFX CAS scaling. Cyberpunk 2077, for example.

And AMD themselves intended to use it together with FSR to reduce its soap.

Using FSR implies EASU + RCAS (FSR being the combination of the two), and generally precludes CAS (older, worse, and less compatible with EASU). Reducing the blurriness of the rather heavy AA (usually TAA) that these upscalers needs to work well, as well as imitate details that didn't exist at the internal resolution, is a major use case for CAS, RCAS, and all these other sharpening filters.

 

[SergMx]

CAS came after the original DLSS and is a pretty basic post process spatial upscaler + sharpen filter that wasn't much different from filters already in existence. It was definitely in the media, and often compared favorably against DLSS 1.0, but it certainly wasn't marketed as well, and was clearly inferior to most DLSS 2.0 implementations.

FSR has been pushed harder and has attracted more attention as it's being presented as a more accessible DLSS near-equivalent.



There are other games using FidelityFX CAS scaling. Cyberpunk 2077, for example.



Using FSR implies EASU + RCAS (FSR being the combination of the two), and generally precludes CAS (older, worse, and less compatible with EASU). Reducing the blurriness of the rather heavy AA (usually TAA) that these upscalers needs to work well, as well as imitate details that didn't exist at the internal resolution, is a major use case for CAS, RCAS, and all these other sharpening filters.

Thank you for your reply. But now I'm not quite sure why both of these modes are in the game. Why include and use the older and worse mode.

 

[Oztig]

Thank you for your reply. But now I'm not quite sure why both of these modes are in the game. Why include and use the older and worse mode.

They took out the 'old' one, and there was an uproar from the forums, so they put it back in, along with the new one.

 

[MacIntoc]

Why include and use the older and worse mode.

Because it can be used to limit the blurring added by many other filters (like AA or VSS).

If i read the doc correctly, CAS can also be used as a dynamic scaler (the rendering resolution is changed on the fly depending, for exemple, on the FPS output). It is this aspect of the technology that FSR makes obsolete.

 

[ebbrell]

I'm finding the scanner shows players as NPCs when NIS is upscaling.
I cannot see who is a player or not on scanner

 

[Morbad]

Thank you for your reply. But now I'm not quite sure why both of these modes are in the game. Why include and use the older and worse mode.

EDO is the only game I've seen both used. However, it's probably prudent to keep the older CAS available as FSR is not universally worse.

One of the guidelines for using FSR is a "well anti-aliased" input...this is not something EDO is able to do currently. The strongest AA EDO has (other than costly supersampling) is a weak SMAA filter. So, right now, there are situations where the older CAS looks better.

I suspect FDev will pull the old method once they sort out their antialiasing problems.

I'm finding the scanner shows players as NPCs when NIS is upscaling.
I cannot see who is a player or not on scanner

NIS does round off text and other fine details.

Does sharpness setting make a difference?

 

[SergMx]

I always thought that the main thing in FSR and DLSS is the number of frames. This is what is always shown when demonstrating these technologies. So that modern computers can pull games in 4k and not slow down.
As for image degradation, it is clear to everyone. There are no miracles, you improve the speed, you worsen the quality.

I have a fairly weak computer R2600 and GF1660 I play on a large TV at a distance from him. I do not care much about the blurriness on the screen as I do not see it much. But I am killed by the drop in fps in 1920 mode.
I tried different ways with FSR, 0 effect. Maybe fsr does not work on nvidia cards ?

Because NIS runs in the driver and FSR in the game. So including both I should get 4x the frame rate increase?

 

[Morbad]

I always thought that the main thing in FSR and DLSS is the number of frames. This is what is always shown when demonstrating these technologies. So that modern computers can pull games in 4k and not slow down.
As for image degradation, it is clear to everyone. There are no miracles, you improve the speed, you worsen the quality.

The degree and precise nature of that degradation varies considerably from algorithm to algorithm and implementation to implementation. Ultimately, the goal is to get the best combination of subjective IQ and objective performance possible.

Personally, I pay much less attention to benchmarks than what the upscalers are doing to the image. Performance is nearly a given...internal resolution, plus a modest hit from whatever overhead is involved, but the precise changes to the image can be all over the place.

I have a fairly weak computer R2600 and GF1660 I play on a large TV at a distance from him. I do not care much about the blurriness on the screen as I do not see it much. But I am killed by the drop in fps in 1920 mode.
I tried different ways with FSR, 0 effect. Maybe fsr does not work on nvidia cards ?

FSR works on NVIDIA cards. One of it's main features is that it's largely hardware agnostic.

Because NIS runs in the driver and FSR in the game. So including both I should get 4x the frame rate increase?

If you can get them to work together (I haven't tried it, but it should be possible) you'll be upscaling an extremely low internal resolution to the base NIS resolution then upscaling it again to the display's native resolution. You'll also be applying two different sharpen filters. I'd expect it to look like crap. It's also going to have more overhead and thus be slower than just picking one or the other and using a lower internal resolution (more performance oriented preset). The only reason to stack them would be if the end result had a visually appealing synergistic effect...I've done this with supersampling and upscalers to improve AA in Elite, but I can't imagine a scenario where using both FSR and NIS would be beneficial.

 

[galahad2069]

Upscaling is a bandaid by Nvidia and AMD to make up for the elimination of SLI/Crossfire, while players play on ever bigger Monitors. FHD is pretty much gone by, 1440p is mainstream, 4K is coming up more and more, a few even want 8K. Graphics cards can't keep up with that, SLI/Crossfire is gone, so upscaling becomes mainstream. Saves R&D cost on actually making hardware that is capable enough to deliver playable FPS.

Graphics cards could perfectly keep up with that in Horizons though, and since the scenes it needs to render are pretty much the same, there is zero reason for Odyssey to be unable to achieve nearly the same performance on the same hardware.

 

[SergMx]

FSR works on NVIDIA cards. One of it's main features is that it's largely hardware agnostic

The translator must not be translating my words correctly. I wrote about my testing, I do not care about quality, but turning on and changing my FSR I do not get an increase in fps speed, why?

I have a frame rate limitation because I'm not interested in frames over 60. But I do not like it when it becomes 5 units. And FSR does not help here.

 

[Morbad]

turning on and changing my FSR I do not get an increase in fps speed, why?

You may not be GPU limited in the areas you are testing.

If the other upscalers, or simply reducing resolution/quality settings, do increase frame rate, but FSR does not, that does indicate a problem with FSR.

FSR is supposed to work on any remotely recent GPU, and does work on my NVIDIA parts going back to at least Pascal.

 

[ebbrell]

I've notice when using NIS, my scanner shows people as solid items on the scanner.

Everyone looks like an npc on the scanner

 

[Agony_Aunt]

I'm finding the scanner shows players as NPCs when NIS is upscaling.
I cannot see who is a player or not on scanner

Is this the solution to ganking we have been looking for? Now all we need to do is convince all the gankers to use NIS with upscaling!

 

[Oztig]

Is this the solution to ganking we have been looking for? Now all we need to do is convince all the gankers to use NIS with upscaling!

Hmm, what res are you using / upscale to, I'm sure I've seen hollow squares and triangles on the scanner

 

[Agony_Aunt]

Hmm, what res are you using / upscale to, I'm sure I've seen hollow squares and triangles on the scanner

Wrong person to ask. You need the person i quoted. Not using scaling.

 

[Oztig]

Wrong person to ask. You need the person i quoted. Not using scaling.

Oops, sorryabout that

 

[SergMx]

I understand NIS in the drivers and that's it.
But the FSR in the game itself, why this setting can not be used dynamically? Turn on and off the FSR and its mode depending on the location?
Inside the station, in the settlement room, the star map, the flight of the FSD, etc.

I transfer from the ship to the SRV screen and so darkens, I leave the SRV in a spacesuit screen also darkens.

 

[Mogen Teras]

Graphics cards could perfectly keep up with that in Horizons though, and since the scenes it needs to render are pretty much the same, there is zero reason for Odyssey to be unable to achieve nearly the same performance on the same hardware.

My assumption is that the switch to a PBR (Physically Based Rendering) system for rendering played a big part in why it performs worse while not looking any better; it's factoring in a ton more variables in how things should look, while doing so within a game engine that was already showing signs of age beforehand.

 

[SergMx]

My assumption is that the switch to a PBR (Physically Based Rendering) system for rendering played a big part in why it performs worse while not looking any better; it's factoring in a ton more variables in how things should look, while doing so within a game engine that was already showing signs of age beforehand.

Many people were surprised that going from a 2D DOOM 1-2 to 3D Quake 1 required the purchase of a Pentium.

 

[galahad2069]

My assumption is that the switch to a PBR (Physically Based Rendering) system for rendering played a big part in why it performs worse while not looking any better; it's factoring in a ton more variables in how things should look, while doing so within a game engine that was already showing signs of age beforehand.

Many people were surprised that going from a 2D DOOM 1-2 to 3D Quake 1 required the purchase of a Pentium.

Oh yes but I guess 3D Quake was looking much better than 2D DOOM, which is not really the case here.
The thing is that whatever PBR is, if all it can do is making the game look worse at the cost of a heavy performance loss, then I cannot imagine a reason why it was worth it to try and implement it in the first place.