Support NeoGAF

[Maxrunner]

More """""""4K"""""" resolution. I'm really not impressed with the PS4Pro.

I found hilarious that comments like come from people who makes indie games with basic 2d retro visuals, just like the guy who made Ori and the Bind Forest.

 

[Wishmaster92]

60fps?

Scorpio!!!

 

[SEGAvangelist]

So dynamic 4K is now basically 4K. OK sony.

It's not dynamic 4K. That would imply that the resolution changes depending on what is currently being rendered.

 

[RoboPlato]

Surprised there's still so much debate about checkerboard rendering in this thread. We know what it is, have seen it's results, and Guerrilla have been upfront about the technique being used.

 

[Sho_Nuff82]

This is completely wrong. For certain frames or portions of frames, CBR is literally indistinguishable from native rendering. In a practical sense, it will often be perceptually indistinguishable.





All this is wrong. There is no "base resolution" for CBR, it's not a form of upscaling. CBR rasterizes just as many pixels as native rendering. Just the method it uses for half of them is slightly less accurate.





For many reasonable definitions of "basically", CBR absolutely is basically 4K. For example, I can 100% guarantee that we can devise fair, honest tests where no one could determine any difference between native and CBR. In real-world applications such perfect results will be scattered, of course. But I think this points out how misguided expectations of giant quality gaps are.


So I assume you refuse to call games "1080p" whenever their shadow buffers, particle effects, or volumetric lighting run at non-native resolution. Without researching, can you name which of last fall's titles aren't "true" 1080p in this sense?

There's a problem with trying to be ideologically pure about things you don't even notice.

You can design fair, honest tests where people can't perceive the difference between 900p, dynamic resolution scaling, and 1080p.

 

[Caayn]

The game will be rendered in 2160p checkerboard (meaning, basically 4K)

The game will be rendered in 2160p checkerboard (meaning, not actually 4K)

I can see why they went with the first one

Yeah. Marketing will have a field day with the checkerboard upscale technique.

 

[Samurai G0SU]

At the end of the day this game will look fan-fuk-intastic! You wonder how much more GG needed to get the game running at a native 4k though.

This will be an amazing addition for the Sony Library. Woo im seatin

 

[Timu]

So it runs at 1920x2160 right? If so, cool.

 

[creatchee]

I mean, if it looks good, is it that big a deal?

 

[RoboPlato]

This is completely wrong. For certain frames or portions of frames, CBR is literally indistinguishable from native rendering. In a practical sense, it will often be perceptually indistinguishable.





All this is wrong. There is no "base resolution" for CBR, it's not a form of upscaling. CBR rasterizes just as many pixels as native rendering. Just the method it uses for half of them is slightly less accurate.





For many reasonable definitions of "basically", CBR absolutely is basically 4K. For example, I can 100% guarantee that we can devise fair, honest tests where no one could determine any difference between native and CBR. In real-world applications such perfect results will be scattered, of course. But I think this points out how misguided expectations of giant quality gaps are.


So I assume you refuse to call games "1080p" whenever their shadow buffers, particle effects, or volumetric lighting run at non-native resolution. Without researching, can you name which of last fall's titles aren't "true" 1080p in this sense?

There's a problem with trying to be ideologically pure about things you don't even notice.

What I posted came from Cerny's interview with DF. I know there's more to it than a standard base res but it was the best answer to the question.

Checkerboarding up to full 4K is more demanding and requires half the basic resolution - a 1920x2160 buffer - but with access to the triangle and object data in the ID buffer, beautiful things can happen as technique upon technique layers over the base checkerboard output.

https://www.google.com/amp/www.euro...tation-4-pro-how-sony-made-a-4k-games-machine

 

[Daffy Duck]

So if it's only circa 4 million pixels v circa 8 million how can that be classed as basically the same?

I don't get it, I've read the thread and seen some comments but it still makes no sense to me.

Maybe it's something I'll just never get.

IDK

 

[Huggy]

Still pretty impressive for a console that has as many bits as a N64.

 

[chico]

im happy with stable 30fps for better visuals

 

[charsace]

QB didn't use checkerboard rendering
Most people disliked the TAA implementation due to ghosting artifacts, not how their super resolution method worked.

It uses what KZSF, Watchdog 2, and R6S uses.

 

[RoboPlato]

So if it's only circa 4 million pixels v circa 8 million how can that be classed as basically the same?

I don't get it, I've read the thread and seen some comments but it still makes no sense to me.

Maybe it's something I'll just never get.

IDK

It's not simply upscaled. Some of that missing data is filled in from samples in previous frames. It's not as sharp as a native image and there are some artifacts but from typical couch distance the differences aren't dramatic.

 

[lyrick]

This is completely wrong. For certain frames or portions of frames, CBR is literally indistinguishable from native rendering. In a practical sense, it will often be perceptually indistinguishable.

"for certain" - There's those fucking weasel words again.

What's the percentage of error from a native 4K image. How many pixels are drawn incorrectly when comparing checkerboarding vs native?

So if it's only circa 4 million pixels v circa 8 million how can that be classed as basically the same?

I don't get it, I've read the thread and seen some comments but it still makes no sense to me.

Maybe it's something I'll just never get.

IDK

Because logically the word 'basically' can be interchanged with the term 'basically not' and not change in actual meaning one bit.

basically 4K = basically not 4K
those 2 phrases are logically the same.

 

[Dee Dah Dave]

So if it's only circa 4 million pixels v circa 8 million how can that be classed as basically the same?

I don't get it, I've read the thread and seen some comments but it still makes no sense to me.

Maybe it's something I'll just never get.

IDK

Because the 4 million 'fake pixels' are taken from the previous frame. Meaning that they are very close to what 'real pixels' would look like. Although a lot depends on how fast the scene is moving.

 

[ktroopa]

It's not dynamic 4K. That would imply that the resolution changes depending on what is currently being rendered.

Pretty sure theres a PS4 Pro ad out there on youtube, that says dynamic 4K gaming. If that has changed well then who cares i guess.

 

[J-Skee]

The debut of the Decima engine was actually with Killzone: Shadow Fall. The only debut is the name itself.

 

[RoboPlato]

It uses what KZSF, Watchdog 2, and R6S uses.

I think SF used a different method.

 

[BRocknRolla]

Pretty sure theres a PS4 Pro ad out there on youtube, that says dynamic 4K gaming. If that has changed well then who cares i guess.

There's plenty of knowledge you could attain about dynamic resolution vs checkerboard vs upscaling. But who cares, right? The only thing I don't get is why someone who doesn't care would care to post about it.

 

[Rellik]

Pixel count is the same. It is rendering a full 3840×2160 framebuffer. Slightly less accurate than Native 4K but still it's basically 4K.



http://www.neogaf.com/forum/showthread.php?t=1330505

This should be first post in all of these threads.

Thanks for this.

 

[Space_nut]

He was probably just checkerbored.

*Basically

 

[Fredrik]

Maybe this will be the game that finally shut up those "PS4 Pro is silent" people.
Sounds like GG is pushing the hardware to the breaking point as usual.

 

[ktroopa]

There's plenty of knowledge you could attain about dynamic resolution vs checkerboard vs upscaling. But who cares, right? The only thing I don't get is why someone who doesn't care would care to post about it.

Welcome to neogaf.

 

[BRocknRolla]

Welcome to neogaf.

I've never understood knowing you're part of the problem, but not having that bother you. I guess that's a topic for another thread.

 

[truth411]

What is the difference between CB 2160p and 2160p native (4k?)??

Also, imagine this engine and zelda.

I'm still confused about checkerboard rendering. But it sounds like checkerboard rendering = 2160i instead of 2160p.

 

[Fliesen]

You can design fair, honest tests where people can't perceive the difference between 900p, dynamic resolution scaling, and 1080p.

2160p checkerboard on a 4k screen and 900p on a 1080p screen are worlds apart in perceptible difference compared to rendering at full native resolution.

2160p checkerboard has a 1:1 mapping between rendered pixels and pixels displayed by the TV.
There is no blurrniess due to an image being scaled from sub native to native.

 

[vpance]

Because the 4 million 'fake pixels' are taken from the previous frame. Meaning that they are very close to what 'real pixels' would look like. Although a lot depends on how fast the scene is moving.

To add to that, pixels on 4K panel are a quarter of the size of those on a 1080p one, so generally the technique achieves far better results at 4K.

Also most TVs still only have poor to ok motion handling (the downside of sample and hold on LCD & OLED). So in fast motion scenes you're not even getting the full resolution of the display unless you use motion flow or whatever mode that adds lag. It becomes more difficult to see CBR artifacts in these situations.

 

[Aceofspades]

Awesome job Gurilla .

 

[Always-honest]

Dear lord it will look insane on my 1080p projector.

 

[geordiemp]

Scorpio!!!

If it has Zen only, otherwise if Jag then games like this or Witcher 3 will always be 30 FPS until Ps5 / Xb2.

Game does look nice, Scopio will also be great, after 1800c....I just cant see differences anymore, its about frame rate.

I'm still confused about checkerboard rendering. But it sounds like checkerboard rendering = 2160i instead of 2160p.

Yeah I think of it like 4KI or interlaced, updated every frame with half of whats needed and using previous frame, framebuffer is full 4K. Looks damn good enough IMO, all games > 1400p I just cant tell anymore on my 4K 55 inch

 

[ktroopa]

I've never understood knowing you're part of the problem, but not having that bother you. I guess that's a topic for another thread.

The problem my friend is 4k and our expectations on pro vs what we being served. See this thread people do care.

 

[Rival]

I don't care what resolution it runs at as it will look great either way. I just hope it runs at a locked 30 fps with perfect pacing. That's all that matters to me.

 

[Liabe Brave]

You can design fair, honest tests where people can't perceive the difference between 900p, dynamic resolution scaling, and 1080p.

I didn't say "perceive". There is no way to devise a fair, honest test where 900p upscaled and 1080p are actually identical. But there are situations where CBR is literally identical to native rendering.

Dynamic resolution is a closer analogue, so then the question becomes a matter of degree. If a game was dynamic but only went below 1080p a small percentage of the time, I'd wager very few people would have a problem with calling it "basically 1080p". It seems to me just as justified, if CBR results are perceivably worse a small percentage of the time, to call it "basically 1080p".

If someone disagrees, they'd need to explain the exact metric on which they base the difference.

 

[DieH@rd]

How can it be basically 4k if it uses half the pixels?

The rest are reconstructed on the fly using data from nearby and past pixels. The end result is a sharp image, not a upscalled blurry one.

Digital Foundry was very satisfied with 2160p checkerboard in RotTR and Infinite Warfare. They described it as very nearly identical to real 4K, with them needing to press they eyeballs to the screen to see which render was real 4K and which was checkerboarded.

When done right, checkerboarding can do wonders.

 

[Koh]

So if it's only circa 4 million pixels v circa 8 million how can that be classed as basically the same?

I don't get it, I've read the thread and seen some comments but it still makes no sense to me.

Maybe it's something I'll just never get.

IDK

My understanding is that the total number of pixels is the same as 4k. Checkerboarding is the "refresh" technique which skips/predicts pixels in a pattern to reduce workload while impacting the image as little as possible.

 

[Shpeshal Nick]

Once Scorpio hits and there are (presumably) multiplatform games that are native 4K on Scorpio and checkerboard 4K on the Pro we can actually see how big the difference is.

Here's the thing.

The difference won't be big. Same way it was never big between Xbox One and PS4. But for 2 years it was made out to be a huge difference when it suited. Now I'm sure until PS5 arrives it won't be a big difference while Scorpio is the slightly more powerful console.

In the end, no one is dual playing these games on a PS4 and Xbox One to see the difference in real time. Each version will look great when it's the only version you're looking at. Same with Pro and Scorpio.

 

[ethomaz]

All this is wrong. There is no "base resolution" for CBR, it's not a form of upscaling. CBR rasterizes just as many pixels as native rendering. Just the method it uses for half of them is slightly less accurate.

Cerny said 1920x2160 buffer for 2160p checkerboard.

Sorry to take Cerny words over yours.

 

[User123456]

In a practical sense, it will often be perceptually indistinguishable.

After perceptual 60fps comes perceptual 4k.

 

[Caayn]

Cerny said 1920x2160 buffer for 2160p checkerboard.

Sorry to take Cerny words over yours.

Cerny is correct. But a checkerboard resolution can't really be expressed in a base resolution as it renders a picture with voids in it. So 1920x2160 is only fit to explain the number of actual shaded pixels (~4.1m), not as the image size which is still 3840x2160 but with holes in it.

 

[Sad Affleck]

The rest are reconstructed on the fly using data from nearby and past pixels. The end result is a sharp image, not a upscalled blurry one.

Digital Foundry was very satisfied with 2160p checkerboard in RotTR and Infinite Warfare. They described it as very nearly identical to real 4K, with them needing to press they eyeballs to the screen to see which render was real 4K and which was checkerboarded.

When done right, checkerboarding can do wonders.

I understand but comparing still images isn't a great way to get an accurate result. Isn't checkerboard rendering supposed to introduce artifacts during motion? What good is a still image comparison? When is the image ever still in a videogame?

 

[Fliesen]

Cerny is correct. But a checkerboard resolution can't really be expressed in a base resolution as it renders a picture with voids in it. So 1920x2160 is only fit to explain the number of actual shaded pixels (~4.1m), not as the image size which is still 3840x2160 but with holes in it.

this. What's the resolution of black squares on an 8x8 chess board.? 4x8? 8x4? ... it's simply "8x8, every other square", just like checkerboard is (in this case) 3840x2160 "every other frame."
It's not a 1920x2160 (thereby 8:9, half width) frame.

I understand but comparing still images isn't a great way to get an accurate result. Isn't checkerboard rendering supposed to introduce artifacts during motion? What good is a still image comparison? When is the image ever still in a videogame?

actually, still images is where you'd be most likely to perceive said artifacts.

The more motion, the more artifcats, the less you'll be able to see them because - well - the image is moving.

 

[KiraXD]

Isn't that a bit close, lol?

nah.. i sit at the end of my bed about 2.5 feet from my 50"

 

[Tripolygon]

Cerny said 1920x2160 buffer for 2160p checkerboard.

Sorry to take Cerny words over yours.

You're misunderstanding what cerny said. There's no base resolution, you start with the native resolution you're trying to render but half filled with traditionally shaded pixels while the other half is filled in from information from previous and current frames.

Take this checkered pattern is 4K but only the white ones are traditionally shaded. The blacks are filled in.

 

[DieH@rd]

I understand but comparing still images isn't a great way to get an accurate result. Isn't checkerboard rendering supposed to introduce artifacts during motion? What good is a still image comparison? When is the image ever still in a videogame?

No, it looks fine in both static and moving scenes. RotTR implementation introduced barelly visible artefacts on very thin objects [like sparks after grendades].

Go look at DF videos about checkerboarding.

 

[ss_lemonade]

Do we really need a PS4 PRO just to get good "anisotropic filtering"? I always thought there was an SDK bug or something that prevented base PS4 games from having a decent amount of anisotropic filtering by default. I guess that isn't really the case now?

 

[Falchion]

If there's one thing these guys can do, it's squeeze every bit of power out of a Sony system.

 

[Kremzeek]

are the tryhards really going to bring up "NOT TRUE 4K!" in every one of these threads?

it's pathetic at this point.

checkerboarding is by its nature an interpolation method to get visuals to look extremely close to native 4K.

and no it's NOT "just an upscale."

 

[Sad Affleck]

No, it looks fine in both static and moving scenes. RotTR implementation introduced barelly visible artefacts on very thin objects [like sparks after grendades].

Go look at DF videos about checkerboarding.

Thanks but it seems I don't understand that stuff at all I've read tons of relevant material and it's all still a complete mystery to me. I think I'll just stop trying and go with what my eyes tell me.