Support NeoGAF

[Zathalus]

As mentioned, I got home and took two screenshots. Both are max settings, with FSR set to 100% in both cases. Frame generation is disabled as well. The only difference is sharpening, one has it enabled at the default setting while the other has it disabled.



I think it's quite clear what sharpening or the lack thereof can do in the end.

 

[Kupfer]

I unsubscribed years ago because I got the feeling they were just influencers and entertainers, doing typical influencer and entertainer stuff to gain attention and, ultimately, money.

However, I still enjoy reading adamsapple ' s summaries of their videos for a quick overview.

 

[ThisIsMyDog]

As mentioned, I got home and took two screenshots. Both are max settings, with FSR set to 100% in both cases. Frame generation is disabled as well. The only difference is sharpening, one has it enabled at the default setting while the other has it disabled.



I think it's quite clear what sharpening or the lack thereof can do in the end.

Imgsli

 

[akira__]

As mentioned, I got home and took two screenshots. Both are max settings, with FSR set to 100% in both cases. Frame generation is disabled as well. The only difference is sharpening, one has it enabled at the default setting while the other has it disabled.



I think it's quite clear what sharpening or the lack thereof can do in the end.

I think it's better if you sit this one out.

FSR is frame generation.

AMD FidelityFX™ Super Resolution 3 (FSR 3)

Discover frame generation with AMD FidelityFX™ Super Resolution 3, and get the source code and documentation!

gpuopen.com

> AMD FSR 3 technology extends upon AMD FSR 2's upscaling by adding Frame Generation – the ability to generate entirely new game frames and present those to the user to improve FPS. FSR 3 does this by adding two new technologies – Frame Interpolation and Optical Flow enhanced from AMD Fluid Motion Frames.

 

[PaintTinJr]

You still didn't prove that the fog is not dynamic.

The fog is dynamic on both, that was the premise of the discussion anyway.

It is the fog precision/quality being garbage on that PC comparison shot that is the issue, and that the two renders are doing different workloads which is visible by the way the fog clumps everywhere between near and far planes on the RX 6700 PC image where geometry fragments at just a pixel or less apart are going from semi opaque to fully opaque, and that's still with the PC shot having regained some fog precision by having the zbuffer values nearest (0.0f) start at a thick fog value (0.5f or more) meaning that the depth precision values using a fog equation converge slower to 1.0f despite the rapid changes of the zbuffer fighting against that.

It is there where we see how poor the precision of the PC RX 6700 fog is compared to the PS5 shot which uses the full (or almost full) range from 0.0f - 1.0f to lightly depth cue in the foreground and vary adjacent pixel opacity with minor changes from foreground all the way to the backplane.

 

[scarlet0pimp]

No they are not biased

 

[Zathalus]

I think it's better if you sit this one out.

FSR is frame generation.

AMD FidelityFX™ Super Resolution 3 (FSR 3)

Discover frame generation with AMD FidelityFX™ Super Resolution 3, and get the source code and documentation!

gpuopen.com

> AMD FSR 3 technology extends upon AMD FSR 2's upscaling by adding Frame Generation – the ability to generate entirely new game frames and present those to the user to improve FPS. FSR 3 does this by adding two new technologies – Frame Interpolation and Optical Flow enhanced from AMD Fluid Motion Frames.

You can independently enable/disable frame generation and use different upscalers in Black Myth. Hence you can enable FSR and not use frame generation at all, as you can in every single other game that has both.

Fucking hell, at least know what you’re talking about before sprouting nonsense.

 

[evanft]

Five edits on a post is wild.

 

[akira__]

You can independently enable/disable frame generation and use different upscalers in Black Myth. Hence you can enable FSR and not use frame generation at all, as you can in every single other game that has both.

Fucking hell, at least know what you’re talking about before sprouting nonsense.

If you and DF would understand upscalers and frame generation, then you would understand why you would want to turn it off for comparisons.

Once enabled each pixel become a guesstimate based on other and previous pixels.

Thus the output no longer is actual output from the produced pixels but interpreted version of them.

 

[adamsapple]

You can independently enable/disable frame generation and use different upscalers in Black Myth. Hence you can enable FSR and not use frame generation at all, as you can in every single other game that has both.

Fucking hell, at least know what you’re talking about before sprouting nonsense.

For your own sanity, stop replying to him lol.

 

[Zathalus]

If you and DF would understand upscalers and frame generation, then you would understand why you would want to turn it off for comparisons.

Once enabled each pixel become a guesstimate based on other and previous pixels.

Thus the output no longer is actual output from the produced pixels but interpreted version of them.

FSR was set to 100% and frame generation was disabled. Hence both upscaling and frame generation were disabled in my images.

For your own sanity, stop replying to him lol.

Good advice, he is likely just trolling at this point.

 

[akira__]

For your own sanity, stop replying to him lol.

But you said the PC version was better in the benchmark, but also that DF was right. Seem you and your buddy Zathalus are in dissagreement over that. He already said that wasnt the case and DF made a oversight.

Still wondering how DF can be right that its the same, but also that the PC version is better according to you. While the images clearly tell another story.

Turning off frame generation and upscaling with FSR meanwhile FSR is frame generation and upscaling. And then say that others are trolling.

 

[evanft]

Six edits in a post is wild.

You can independently enable/disable frame generation and use different upscalers in Black Myth. Hence you can enable FSR and not use frame generation at all, as you can in every single other game that has both.

Fucking hell, at least know what you’re talking about before sprouting nonsense.

Yeah it's kind of weird he didn't know this. I haven't seen a game yet that didn't have them as independent options.

 

[akira__]

Six edits in a post is wild.



Yeah it's kind of weird he didn't know this. I haven't seen a game yet that didn't have them as independent options.

You can't turn off upscaling and frame generation and use FSR. That is the point of FSR.

 

[Zathalus]

You can't turn off upscaling and frame generation and use FSR. That is the point of FSR.

You can. It’s pretty clear you are not familiar with PC gaming. FSR (and DLSS) is three things, an upscaler, a method of AA, and frame generation. If you disable frame generation (with the in game toggle in the menu) and you set FSR to 100% you disable the upscaling part as the 100% refers to what percentage of the native resolution should the game upscale with. 100% obviously means native resolution so FSR (and DLSS) is just used as an AA method.

Most PC gamers are aware of these settings, it’s not some arcane wizardry.

 

[akira__]

AA generated by... ML and AI. thus we are back to starting point. That is the difference between normal AA and FSR. It's how it's done, is the important difference. By incorporating ML and AI you could very easily get different benchmarks especially with still images.

AA done by FSR works because they trained ML models to predict how a sharper image would look like. Issue is that it can create weird artifacts.

Where as "normal" AA is based of algorithms.

 

[Zathalus]

AA generated by... ML and AI. thus we are back to starting point. That is the difference between normal AA and FSR. It's how it's done, is the important difference. By incorporating ML and AI you could very easily get different benchmarks especially with still images.

AA done by FSR works because they trained ML models to predict how a sharper image would look like. Issue is that it can create weird artifacts.

Where as "normal" AA is based of algorithms.

FSR doesn’t even use ML.

 

[akira__]

FSR doesn’t even use ML.

unfortunately for me that' correct
went into technical papers and they are not using ML, but are using the lanczo algorithm.

Lanczos algorithm - Wikipedia

en.m.wikipedia.org

Since nvidea had those technologies first I presumed they worked the same.

That said, they both create weirdness during benchmarking since the image still are interpreted.

But since the ps5 pro uses it, we can set our watch, because DF will find all the issues with framegen and upscaling no worries.

And then in a couple of week from now you will make arguments against using frame gen, upscaling.

 

[Eszti]

It took Digital Foundry 4 years to simply raise this question:

no, they live out of content its that simple. they are not stupid, they are maximizing profits.

 

[PaintTinJr]

As mentioned, I got home and took two screenshots. Both are max settings, with FSR set to 100% in both cases. Frame generation is disabled as well. The only difference is sharpening, one has it enabled at the default setting while the other has it disabled.



I think it's quite clear what sharpening or the lack thereof can do in the end.

Putting aside that your FOV choice massively alters too many things in the rendering to be representative of fixing an issue in an image with a different FOV, neither of those images are superior to the other in detail numerically when analysed, and just comparing the histograms, all reshade is doing from its process is altering the gamma curve to give the second image a wider range which in turn improves clarity without adding detail. The softer blurry edges in the first image are lost because neighbouring objects then appear to be separately lit giving better contrast.

Doing image 1 - 2, or 2 - 1, yields crunchy results either way round, so reshade isn't improving the image technically, unlike the clear difference in actual detail between the PS5 balance images and DF RX6700 PC, so it isn't that either.

 

[SKYF@ll]

No, they don't.

Maybe in screenshots in photo mode. Hell, even with DLSS some few games IQs don't look great.

I have just captured the same scene as the image you presented. (PS5 Performance Mode)
Image compression is similar. * your SS:1080p(230KB) vs my SS:2160p(1MB)

 

[adamsapple]

I have just captured the same scene as the image you presented. (PS5 Performance Mode)
Image compression is similar. * your SS:1080p(230KB) vs my SS:2160p(1MB)

The game has been patched to hell and back now, and has removed the forced RT in Performance Mode, which was the biggest performance hogger at launch.

So it would be impossible to get the exact same comparison now that represents how the game looked at launch.

 

[Physiognomonics]

Where did the developer say that the texture settings are the same?
There is a significant difference in image quality just from the difference in sharpness.
These images were captured by me previously. (I don't have the PC version.)

My take on this it is simply lowered resolution textures on XSX because of memory constraints. Yes on XSX too.

 

[Gaiff]

You can't turn off upscaling and frame generation and use FSR. That is the point of FSR.

AA generated by... ML and AI. thus we are back to starting point. That is the difference between normal AA and FSR. It's how it's done, is the important difference. By incorporating ML and AI you could very easily get different benchmarks especially with still images.

AA done by FSR works because they trained ML models to predict how a sharper image would look like. Issue is that it can create weird artifacts.

Where as "normal" AA is based of algorithms.

Pretty clear as day you’re just here to be a brick wall. You didn’t even do basic research. Frame Generation wasn’t introduced until years into FSR’s life so the very thought that we cannot decouple them is beyond moronic when FSR has existed before AMD frame generation was even a thing.

Doesn’t matter what you’re told, you’re going to shift the goalposts. Didn’t even basic details such as FSR not using ML.

Also lol at going into a "technical" paper. Shows how much you know when you weren’t even aware FSR isn’t using ML or being independent from frame generation.

 

[PaintTinJr]

Pretty clear as day you’re just here to be a brick wall. You didn’t even do basic research. Frame Generation wasn’t introduced until years into FSR’s life so the very thought that we cannot decouple them is beyond moronic when FSR has existed before AMD frame generation was even a thing.

Doesn’t matter what you’re told, you’re going to shift the goalposts. Didn’t even basic details such as FSR not using ML.

Also lol at going into a "technical" paper. Shows how much you know when you weren’t even aware FSR isn’t using ML or being independent from frame generation.

Disregarding someone completely on the basis of FSR not being technically AI upscaling and not always having frame-gen, despite the version of FSR being used here has the feature seems off, and certainly doesn't undermine the premise of the thread or other observations the OP made.

Zathalus has repeatedly said Reshade makes the RX6700 based DF PC image the same as the PS5 balance, and after him putting up images to analyse, it doesn't add in missing detail, but just alters clarity via gamma and tones, and his claim was wrong but no one is going to say on that basis that all of his other comments in here are invalid, so a bit of wiggle room seems necessary for all. And on the FSR mistake, it does still work wonders and lives in the same bracket of tools as DLSS and XeSS, CB, TAA, and is always held to the same standards as DLSS by gamers and DF because gamers use it in the same way to rebalance image quality and performance, so is hardly a massive mistake.