Microsoft: "Tiled Resources" Key To Xbox One Graphics

[Somebloke]

Given that I doubt the Kinect camera can manage all that much eye-tracking precision, with several metres between the user and the screen and likely dark rooms, I am going to assume this will mostly be a matter of concentrating detail around contextual things like the centre of the screen, for first person views, or around the cursor if the title has one, or in places to which the game designers figure themselves attracting viewer attention, e.g. because that's where the action takes place.

Will it be annoying to not get to choose what part of the screen you want to inspect closer, at high LOD level, without having to move in-game focus over there (EDIT: ...such as with the highly irritating depth of field effects you may have previously seen in games), or will we never notice? What about onlookers, if the player is not alone? -Time will tell, I guess. :7

Wot. They're talking about the game's field of vision, not your eyes. Basically, lets say you're playing an FPS, the game only needs to render everything in your field of vision. You can't see anything behind you so the game saves resources by not loading them

Geometry outside your field of view, as well as occluded parts, have always been culled in 3D rendering, ever since people started doing it.

This is about switching to lower detail rendering (...such as models with fewer polygons, reducing the number of texture layers (not changing to smaller bitmap sizes - that offers no to little speed gain; I notice some people seem to misunderstand the concept of mipmapping), using simpler particle systems, etc...) for things that are on screen, based on criteria in addition to the usual distance-based Level-Of-Detail complexity reduction.
The idea is that if you are looking where you aim, eyes on the crosshair, stuff towards the edge of the screen will be in your peripheral vision and can be rendered with less detail, because you are not likely to notice that badly.
The fovea of the human eye is really tiny.

 

[Voltano]

Correct me if I'm wrong but don't they basically already do this in games? I think they call it occlusion culling? seems similar, at least to me.

It sounds similar, but it doesn't seem like they are the same. "Occlusion Culling" seems to only hide parts of a mesh's sides from a specific point of view. A plane GameObject in Unity 3D is a good example where you can see one side of the plane from one direction, but the other side of the plane is invisible. What this is doing is making objects appear blurry or use lower-quality textures/low poly models when they are not in "focus" of the game's camera. And they use higher textures/more poly models when the camera does have focus on them. I think.

For me, this won't make me pick up an Xbox One. The technology doesn't sound new to me as YoYo Games' "Game Maker 8.1" is capable of disabling objects off-camera to make up for memory space. Even then this is only reducing the quality of visual effects on objects that are not in focus of the camera, so I doubt this would be used to make a game have better core mechanics.

Off-topic: Would this mean that instead of looking at these games now with a camera, we're watching our avatars through someone's eyes with this analogy? Would that mean if they give us an HD version of "Kane & Lynch 2" with this technology, does that mean some unknown person is looking at a bunch of old, naked men in a suburban area? Ew...

 

[Ryan Hughes]

Do you have a source for this? All I've been able to find are stories that lead back to the same article.

Anyway, hopefully texture popping won't be as big an issue now that we have plenty of RAM to play with, since they should be able to keep more of it in memory.

It is common knowledge. Texture rendering such as this has been going on for almost ten years now. Unreal Engine 2 used it, which was built on Microsoft's Direct 3D at the time. This is why some games using Unreal Engine pop their textures in and out.

Also, the Xbox One will not have that much memory, because its three operation systems already take up 3gb off the top. Overall, this is just a way of refining an old technology to make up for their own asinine use of system specifications.

 

[BloodSquirrel]

It is common knowledge. Texture rendering such as this has been going on for almost ten years now. Unreal Engine 2 used it, which was built on Microsoft's Direct 3D at the time. This is why some games using Unreal Engine pop their textures in and out.

Also, the Xbox One will not have that much memory, because its three operation systems already take up 3gb off the top. Overall, this is just a way of refining an old technology to make up for their own asinine use of system specifications.

That's not "common knowledge", that's "making assumptions about what the tech does".

I realize that it's not common around here, but I'd like a real basis from which to speak about whether this tech is significant or not before rendering an opinion on it.

 

[Lunar Templar]

Yeah? And? Sure it's a lovely and effective skill an all but it doesn't matter how pretty it looks if there's nothing worth looking at.

 

[Angelous Wang]

Basically the way texture swapping now works the game has two or three textures, one crappy one for when you are far away from the object, good one when you are up close and sometimes one less crappy one for some pre-set intermediate distances. The game swaps between the textures in 360 (every direction) radius around the character based on position and distance.

What this "new" version is doing is only swapping the actual on screen (and still only close objects) for the good ones. Which all in all means a lot less of the game will be using the good textures at once.

Of course the main problem with this would still be the one we have now with normal texture swapping, texture popping. Because textures are going to be swapping around so much more all over the place, which means if you get any slow texture swapping textures will be popping every time you turn around.

Do you have a source for this? All I've been able to find are stories that lead back to the same article.

Anyway, hopefully texture popping won't be as big an issue now that we have plenty of RAM to play with, since they should be able to keep more of it in memory.

It's in the source for the article.

Tiled resources automatically swaps in the details whenever a gamer is looking closely at it.

Leblond demonstrated this technology using a map of Mars that was blurry whenever he zoomed into the landscape. He then turned on tiled resources, which made the minutiae of the surface of Mars pop with crispness.

 

[Phrozenflame500]

The xBox One and Windows 8.1, now with blast processing tiled resources!

 

[DiamanteGeeza]

Of course the main problem with this would still be the one we have now with normal texture swapping, texture popping. Because textures are going to be swapping around so much more all over the place, which means if you get any slow texture swapping textures will be popping every time you turn around.

You're confusing two problems. Mipmapping doesn't 'pop' textures - the GPU blends between the two current levels of mip so the transition is smooth. I would assume this will do the same.

The 'popping' you're talking about is because of streaming data. Typically, what we do is ensure that the lowest detail texture for every texture in the level (the lowest mip, typically, but what I'm about to explain is not mipmapping) is always resident in memory so there will always be some sort of texture there, and the higher res textures stream in as needed. In an ideal world, all high res textures will be there when required, but all manner of things can screw this up for a few frames; typically what you're seeing is the disk read speed just not being able to keep up with the amount of data that's being demanded from it, so you'll sometimes be looking at the lowest level mip of certain textures until the higher res texture has finished loading and it 'pops' into view.

 

[Pyrian]

The idea is that if you are looking where you aim, eyes on the crosshair, stuff towards the edge of the screen will be in your peripheral vision and can be rendered with less detail, because you are not likely to notice that badly.
The fovea of the human eye is really tiny.

The problem with this notion is that peripheral vision is quite sensitive to motion, and so will very easily pick up on the textures changing as you look around. A lot of games which look fantastic in screenshots, look very strange in motion, because, as others have said, this sort of thing is hardly new.

 

[BloodSquirrel]

Tiled resources automatically swaps in the details whenever a gamer is looking closely at it.

Leblond demonstrated this technology using a map of Mars that was blurry whenever he zoomed into the landscape. He then turned on tiled resources, which made the minutiae of the surface of Mars pop with crispness.

I've read that. It neither defines what it means by "details" nor "whenever a gamer is looking closely at it". The guy who wrote that sentence isn't even a technical expert- he's a just a journalist, a profession which has a history of giving horribly inaccurate summaries of new technologies and scientific discoveries.

 

[Frezzato]

Correct me if I'm wrong but don't they basically already do this in games? I think they call it occlusion culling? seems similar, at least to me.

You just brought back good memories of the map maker in Far Cry 2

Conceptually, I'm more interested in the cool tools that other companies might bring to the next gen, like Criterion's Renderware, which allowed both Burnout 3 and GTA San Andreas to run on both PS2 and original Xbox.

 

[RicoADF]

Correct me if I'm wrong but don't they basically already do this in games? I think they call it occlusion culling? seems similar, at least to me.

It sounds similar, but it doesn't seem like they are the same. "Occlusion Culling" seems to only hide parts of a mesh's sides from a specific point of view. A plane GameObject in Unity 3D is a good example where you can see one side of the plane from one direction, but the other side of the plane is invisible. What this is doing is making objects appear blurry or use lower-quality textures/low poly models when they are not in "focus" of the game's camera. And they use higher textures/more poly models when the camera does have focus on them. I think.

For me, this won't make me pick up an Xbox One. The technology doesn't sound new to me as YoYo Games' "Game Maker 8.1" is capable of disabling objects off-camera to make up for memory space. Even then this is only reducing the quality of visual effects on objects that are not in focus of the camera, so I doubt this would be used to make a game have better core mechanics.

Off-topic: Would this mean that instead of looking at these games now with a camera, we're watching our avatars through someone's eyes with this analogy? Would that mean if they give us an HD version of "Kane & Lynch 2" with this technology, does that mean some unknown person is looking at a bunch of old, naked men in a suburban area? Ew...

Rage had the tiling system where it blured what wasn't in the centre of view. It didn't look good and more often than not attracted the attention of eyes due to it being 'movement'. So it isn't new nor is it a good thing.

 

[Somebloke]

The problem with this notion is that peripheral vision is quite sensitive to motion, and so will very easily pick up on the textures changing as you look around. A lot of games which look fantastic in screenshots, look very strange in motion, because, as others have said, this sort of thing is hardly new.

Not saying it is - it isn't. However; most of the reasons I see given in this thread are the wrong ones, stated out of misunderstanding. Also: do not make the assumption that you and I are in opposition.

The transitions between detail levels, that you see, are going to be similar to the ones you see in a far away object, in a game today, as you close in on it -- only this time they will take up a significantly larger part of your field of view and, for that reason, as if there weren't others, be more noticeable.
...and definitely noticeable, should you have the audacity to not be looking at the part of the screen that the game expects you to.

As far as motion goes: The object transformations will be exactly the same, regardless of the amount of detail they are rendered at and that is the most significant type of motion that the brain can't let slide, while I could easily picture myself not noticing more "subtle" animations being simplifed.

How abrupt the changes in appearence will look, is dependent on how well the LOD levels are designed and what additional effects (such as crossfading, motion blur, DOF, meshes possibly being hardware tesselated to begin with, etc) may be used to mask the swap.

 

[thesilentman]

... This still doesn't tell me anything I'd want in a console. Fuck that, it doesn't tell me anything about how the tech works. How it really works.

Still not getting one, as I'm happy with a future Wii U and a PC due to guess what? The games. Don't bother selling me your console, MS, if you're not going to make it appealing. -.-

 

[CriticalMiss]

Did anyone else read the headline as 'Microsoft: "Tired Resources" Key To Xbox One Graphics'? From some of the comments here it doesn't sound like it is a revolutionary new technology and it also bring in to question how good the hardware is if they need to chop out some of the textures off screen to make the games look good.

 

[Demandred20]

Is tiled resources shorthand for a PC with Intel chips and Nvidia graphics card?

 

[oldtaku]

Well, it's cool to have that standardized in DX, but you'd be crazy to buy any sort of console just because of that.

 

[deadish]

Not exactly high on details here ...

Not even sure it's a good idea. From what I can tell, they are suggesting developers render the edges of the screen at lower quality than the centre.

Don't see why this isn't possible on other consoles to be honest.

Edit: After a bit of research ... "Titled Resources" ... Isn't this what id Software's Rage does? AKA MegaTexture. We already had that on the PS3 and 360 - granted with the terribly small amount of memory it didn't work too well.

So MS is basically asking developers to work around it's lack of memory bandwidth on their main 8GB of RAM (and the lack of space on the 32 MB ESRAM) by streaming textures?

 

[DSK-]

Old technology is old. I just found this, and it explains it (well, mipmapping that is) rather well, for those who aren't aware of what it does.

http://luki.webzdarma.cz/eng_08_en.htm