• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

How do you explain checkerbeard rendering in as few words as possible?

Yes, but they're not all germane. You have a particular hair color, and personal biases about rendering techniques, and living relatives in particular parts of the world, and a certain level of visual acuity, and.... Plenty of that stuff is unimportant; you're naturally editing every response you make to me, cutting away enormous swaths of context in order to sharpen your argument to its cogent leading edge.

For your position to stand, what's required isn't for you to justify the truth of what 1080p-->2160p nearest-neighbor upscaling looks like. What's required is for you to justify why 2160c must only be compared to 1080p-->2160p nearest-neighbor upscaling. You don't seem to have even attempted this, just repeatedly asserted it must be so.


But that difference has nothing to do with a particular amount of upscaling. It lies in the method, as can be seen by the responses everyone has given in this thread. The GIF is from a thread with a different goal, so it isn't a description of what CBR is doing. It's an illustrative example of how its results might differ from "correct" rendering, compared to upscaling of the same game on the same machine.


Actually, it seems many (most?) people do not. Even in game mode, almost all displays do not use nearest-neighbor for their upscale. (The same is true for hardware scalers built into GPUs.) From some casual searching the world brain, it seems only recent 4K TVs have started to add this option, and sometimes as its own mode (not in Game mode). The great majority of displays do not, and use other methods for upscaling at all times. For now, it seems the comparison you want to focus on actually occurs quite rarely.


I don't think you understand the point of the balancing. Yes, each device has a different power profile, but the comparison isn't meant to distinguish between an uber rig and a potato. It's instead between similar software running on the same hardware. Obviously, the same amount of performance is available, so CBR will be up against an upscaled regime that closely matches it in overall demand placed on the hardware.


This is a very odd thing to say. I wouldn't even begin to characterize the comparison of CBR and upscaling in the original GIF as "parity". They're considerably different results...which I thought you also recognized, given your comment that the upscaling looked "like complete shit". I apologize, but I can't really parse what your complaint is now. Does 1800p-->2160p bilinear upscaling look too good? Does it look too bad? What problem are we attempting to resolve?
OK, just to be clear. This is what the OP was addressing.

So when this talk comes out with some of my friends they often make ridiculous remarks as "PS4 Pro games are not native 4K, so the Xbox One S can upscale their games to 4K too."

So, showing the image with the upscaled 1080p image is relevant to the conversation. I'm not sure why you keep trying to steer it away from that and putting words in my mouth just so you can argue against them. You keep bringing up arguments that have no context to my statement or the OP's question. What does 1800p-->2160p bilinear upscaling have to do with the Xbox One S 900p or 1080p upscaling? It's like you are intentionally ignoring context.
 
TL/DR: It's an advanced form of interlaced rendering with an even more advanced deinterlacing algorithm.

Each frame renders an alternating half of the image, just like interlacing, but using a checkerboard pattern instead of alternating horizontal lines.

A static image resolves into a full 4k picture after two frames, just like interlacing forms a full progressive frame from a static image. Still camera screenshots of CB rendering look indistinguishable from native 4k for this reason.

If that as all CB did, there would be checkerboard artifacts whenever something moved on the screen, just like raw interlacing causes combing artifacts. To avoid that, a buffer containing motion vectors is used to find out where the missing pixels were in the previous frame. Since that information isn't always available, some blending is used which can cause a slight loss of resolution on certain moving parts of the frame.

So no, it's not like upscaling at all.
 
That's an accurate image. Some facts are just reality. This thread is about explaining​ the difference in the simplest terms possible. That's it. That's what people see on their 4K screens when a 1080p image is upscaled in game mode without all the lag producing TV settings like Sony's reality creation trying to upres it on the fly. It's​ not supposed to represent the same amount of rendering work. The Xbox One S, PS4, PS4 Pro, and Xbox Scorpio aren't doing the same amount of rendering work. This thread is about explaining the difference between what people will see, not trying to create some artificial parity as a talking point.
I only know of a single television which offers a pixel doubling/nearest neighbor option for scaling 1080p to 4K, and it's disabled by default.
It's also an odd-duck in that it was a TV with a DisplayPort connection for 4K60 back when HDMI was limited to 4K30.
dsc00881j0s4ma7utl.jpg

Virtually all displays use some form of filtered upscaling.
It's actually something that I wished more displays had as an option.

If you use nearest neighbor scaling, the image appears just as though you were viewing it on a 720p/1080p native panel instead of having a somewhat blurred appearance.
Whether that's an improvement or not depends on what you're trying to view.

Additionally, many people have their console handle the scaling instead of the display.
 
Best way to describe the difference is that upscaling just stretches a smaller resolution over a bigger canvas and doesn't create any new information. The image is still going to be the same resolution as before, just with 4 pixels per original pixel.
Checkerboard rendering actually gives you a 4k picture. Each pixel is different to its neighbouring pixel. Half of these pixels are rendered normally, the other half is rendered using programming magic. Unlike upscaling, you're getting 4k worth of image instead of just a stretched canvas.
 
Best way to describe the difference is that upscaling just stretches a smaller resolution over a bigger canvas and doesn't create any new information. The image is still going to be the same resolution as before, just with 4 pixels per original pixel.
Checkerboard rendering actually gives you a 4k picture. Each pixel is different to its neighbouring pixel. Half of these pixels are rendered normally, the other half is rendered using programming magic. Unlike upscaling, you're getting 4k worth of image instead of just a stretched canvas.

The other half isn't rendered by "programming magic", they come from the previous frame because the checkerboard pattern alternates every frame like in interlaced video.


The alternating pattern is a key part of the process that puts it in a different category than "upscaling by definition".
It's upscaling by definition
image.php


(just for kicks: open https://en.wikipedia.org/wiki/Interlaced_video and search for "upscale" or "upscaling" and see that it returns zero results)
 
OK, just to be clear. This is what the OP was addressing.
And note that he already understands that comparing 1080p upscaling to CBR is "ridiculous". Many ways to hurdle that objection are readily available. But the most effective tactic for undermining an opponent's argument is to show inescapable flaws even in the most generous interpretation of their position.

That's what the original GIF does: grants the upscaling method the nicest possible starting point--far better than simple 1080p squaring--and shows that CBR is superior even to that. Worse use cases are demolished as a matter of course.

I only know of a single television which offers a pixel doubling/nearest neighbor option for scaling 1080p to 4K, and it's disabled by default.
It's also an odd-duck in that it was a TV with a DisplayPort connection for 4K60 back when HDMI was limited to 4K30.

Virtually all displays use some form of filtered upscaling.
It's actually something that I wished more displays had as an option.
At least some very recent Sony TVs also have a nearest-neighbor mode, intended to display 2D graphics optimally (given the usual presence of small, sharp detail). But you're right that nearest-neighbor is generally quite rare in displays, and pretty much nonexistent in GPUs.

It's upscaling by definition
If this is the case, it should be easy for you to identify which step in CBR is a scaling operation. Can you do that?
 
Upscaling is just blowing up an image, like you do with a projector moving it away from the wall. It changes nothing in the image quality which means more blurriness and aliasing the bigger it gets.

Checkerboard is a whole new rendering process. Half the pixels are rendered in native 4k, the rest is filled in.

In many cases, it's almost indistiguishable from native 4k while using half the resources according to some outlets like digital foundry and such that do comparisons.

It will be improved upon and used more and more as we get to 8k and beyond because rendering native is a waste of resources according to most devs who would like to use the extra resources to improve filters, particles and graphics in their games.
 
The other half isn't rendered by "programming magic", they come from the previous frame because the checkerboard pattern alternates every frame like in interlaced video.


The alternating pattern is a key part of the process that puts it in a different category than "upscaling by definition".
image.php


(just for kicks: open https://en.wikipedia.org/wiki/Interlaced_video and search for "upscale" or "upscaling" and see that it returns zero results)
I was purposefully dumbing it down. He wants a simple explanation for why it's different to upscaling. Doesn't need to know HOW it generates the extra pixels.
 
And note that he already understands that comparing 1080p upscaling to CBR is "ridiculous". Many ways to hurdle that objection are readily available. But the most effective tactic for undermining an opponent's argument is to show inescapable flaws even in the most generous interpretation of their position.

That's what the original GIF does: grants the upscaling method the nicest possible starting point--far better than simple 1080p squaring--and shows that CBR is superior even to that. Worse use cases are demolished as a matter of course.

Yes, of course OP already understands, that's not what he was asking. He was looking for the simplest way to explain it to people who don't understand what upscaling even is. The original GIF requires too much explanation. To the point that the people he's talking to would likely lose interest or just scoff at his attempt to explain it. He needs simple and to the point. That's why a quick and dirty 4:1 is the best answer in his case. You get too complex with someone who doesn't understand what you're talking about and your argument actually loses credibility with them because it doesn't mean anything to them.

He needs a way to dumb it down Barney style.
 
The other half isn't rendered by "programming magic", they come from the previous frame because the checkerboard pattern alternates every frame like in interlaced video.

The alternating pattern is a key part of the process that puts it in a different category than "upscaling by definition".
image.php


(just for kicks: open https://en.wikipedia.org/wiki/Interlaced_video and search for "upscale" or "upscaling" and see that it returns zero results)

It is rendering a subset of the pixels and filling out the rest in order to display in a higher resolution. The fact that it can use previous frame information doesn't change that.

Are you going to tell me temporal AA isn't AA? Probably not.

It's just semantics, I guess when some people say upscaling they don't mean it in the literal sense, they mean it in a more narrow sense.
 
Yes, of course OP already understands, that's not what he was asking. He was looking for the simplest way to explain it to people who don't understand what upscaling even is. The original GIF requires too much explanation. ...He needs a way to dumb it down Barney style.
"This GIF shows CBR is much closer to native than really good upscaling. Xbox One S upscaling is worse." That seems pretty darn simple to me.

I really think you'd get at least as much if not more pushback and confusion with the updated version, where the 1080p squaring looks totally unrelated (and you'd have to explain that it's probably not what they'd actually be seeing).

It is rendering a subset of the pixels and filling out the rest in order to display in a higher resolution. The fact that it can use previous frame information doesn't change that.

Are you going to tell me temporal AA isn't AA? Probably not.
No, but I will tell you that it isn't "lighting", which is the same sort of illogical redefinition of terms you're pursuing when you say CBR is "upscaling".

It's just semantics, I guess when some people say upscaling they don't mean it in the literal sense, they mean it in a more narrow sense.
Yes, they mean it in a more narrow, truly literal, actually correct sense. As opposed to you saying "Every single way of filling in pixels is upscaling!", which is the opposite of the literal meaning.

The statements correcting you are "just semantics" in the same way that insisting cars are not planes is "just semantics".
 
"This GIF shows CBR is much closer to native than really good upscaling. Xbox One S upscaling is worse." That seems pretty darn simple to me.

I really think you'd get at least as much if not more pushback and confusion with the updated version, where the 1080p squaring looks totally unrelated (and you'd have to explain that it's probably not what they'd actually be seeing).


No, but I will tell you that it isn't "lighting", which is the same sort of illogical redefinition of terms you're pursuing when you say CBR is "upscaling".


Yes, they mean it in a more narrow, truly literal, actually correct sense. As opposed to you saying "Every single way of filling in pixels is upscaling!", which is the opposite of the literal meaning.

The statements correcting you are "just semantics" in the same way that insisting cars are not planes is "just semantics".
We'll just have to disagree, then. Look, I understand what you're saying, but random person on the street wouldn't look at it the same way. In order to use the first image you have to assume random person on the street would understand what you are trying to say with it. I don't have that kind of faith in them. You need to dumb it down to "look, fewer squares."
 
Top Bottom