GTS vs GT6 - Cars comparison

  • Thread starter emula
  • 1,346 comments
  • 277,702 views
eran0004
*Adaptive tessellation.

All 3D meshes have a tessellation. The difference with adaptive tessellation is that it adapts the tessellation depending on the circumstances in the scene, such as the distance to the camera and how busy the scene is.
Click to expand...
So in short it's dynamic and adjust to better allocate resources depending on situations? I feel like the term has been misused many times since its introduction. Would it be correct to point it out as a thing when it's just a single model on a scape? As I've seen people do as such, and continue to berate and put down other games models graphical beauty because of it, when they're comparing single car shots with nothing else around or going on.
 
ImaRobot
So in short it's dynamic and adjust to better allocate resources depending on situations? I feel like the term has been misused many times since its introduction. Would it be correct to point it out as a thing when it's just a single model on a scape? As I've seen people do as such, and continue to berate and put down other games models graphical beauty because of it, when they're comparing single car shots with nothing else around or going on.
Click to expand...

It’s always a thing, and it probably makes the models look better at any circumstances (since adaptive tesselation increases model resolution where it matters). The best rendering quality of course is in photo mode, thanks to the longer rendering time, but for scapes in particular it’s mainly the HDRI environment map (basically the photograph used as a scene) that makes the models look so realistic, due to the way it provides highly realistic lighting and reflections.

Zooming in on details do benefit from adaptive tessellation, since there is no limit on the model resolution. With fixed tessellation the model is rigid and doesn’t improve as you zoom in, so eventually you will notice some rough edges.

The best use of adaptive tessellation is during gameplay though, since the rendering time is so short. Surfaces near the camera can be rendered with a higher resolution than surfaces further away, and that’s a more efficient use of the resources.
 
eran0004
It’s always a thing, and it probably makes the models look better at any circumstances (since adaptive tesselation increases model resolution where it matters). The best rendering quality of course is in photo mode, thanks to the longer rendering time, but for scapes in particular it’s mainly the HDRI environment map (basically the photograph used as a scene) that makes the models look so realistic, due to the way it provides highly realistic lighting and reflections
Click to expand...
But wouldn’t the scapes mode basically be a starting point from what it would start adjusting, adjusting downward from there, rather than adjusting upward from another point? You’d think that would be the default point from where the adjusting would start, no?

eran0004
Zooming in on details do benefit from adaptive tessellation, since there is no limit on the model resolution. With fixed tessellation the model is rigid and doesn’t improve as you zoom in, so eventually you will notice some rough edges
Click to expand...
ah ok, this clears it up for me, somewhat. It seems to be a hit and miss situation for some vehicles.

eran0004
The best use of adaptive tessellation is during gameplay though, since the rendering time is so short. Surfaces near the camera can be rendered with a higher resolution than surfaces further away, and that’s a more efficient use of the resources.
Click to expand...
Yeah that makes sense, as it would be less demanding to run it in photo mode/scapes I would imagine, and the allocating of resources during gameplay will make the most use of it for the models there. Thanks for the clarification.
 
ImaRobot
But wouldn’t the scapes mode basically be a starting point from what it would start adjusting, adjusting downward from there, rather than adjusting upward from another point? You’d think that would be the default point from where the adjusting would start, no?
Click to expand...

Tessellation makes it smoother, so it actually adjusts upwards. It might be able to adjust downwards as well, but I don’t know about that.

So when modelling a car, you don’t model all the perfectly smooth surfaces down to the millimetre, instead you model a fairly rough approximation and then use tessellation to make it smoother (and a smooth shading algorithm as well). With tessellation, you can start with a cube (one vertice in each corner) and end up with a perfect sphere.

Below is an example of the Catmull-Clark algorithm, starting with a cube:

220px-Catmull-Clark_subdivision_of_a_cube.svg.png
 
@emula would you be open to take the pictures of the really expensive cars over shareplay? The taken photos can be extracted via usb and mailed to you so you can post them.
I have the jaguar, Ferrari and the Lamborghini and would be willing to help.
Edit: and I have a PS4 pro, so 4k png pictures are technically possible
 
oinksnork
@emula would you be open to take the pictures of the really expensive cars over shareplay? The taken photos can be extracted via usb and mailed to you so you can post them.
I have the jaguar, Ferrari and the Lamborghini and would be willing to help.
Edit: and I have a PS4 pro, so 4k png pictures are technically possible
Click to expand...

Thank you, I've already tried this method for the 330 P4 comparison

https://www.gtplanet.net/forum/threads/gts-vs-gt6-cars-comparison.370800/page-16#post-12213626

but I prefer take the photos in both games for a better comparison
 
eran0004
Tessellation makes it smoother, so it actually adjusts upwards. It might be able to adjust downwards as well, but I don’t know about that.

So when modelling a car, you don’t model all the perfectly smooth surfaces down to the millimetre, instead you model a fairly rough approximation and then use tessellation to make it smoother (and a smooth shading algorithm as well). With tessellation, you can start with a cube (one vertice in each corner) and end up with a perfect sphere.

Below is an example of the Catmull-Clark algorithm, starting with a cube:

220px-Catmull-Clark_subdivision_of_a_cube.svg.png
Click to expand...
The "downward" adaptation was also used on some cars in GT6, but is not easily do-able on PS4 hardware because of the lack of programmability in the vertex stream processors (compared to the PS3 / Cell's SPUs).


Anyway, it's known as a progressive mesh :)

You start with the highest resolution mesh (as modeled, before subdivision, as you describe) and then simplify it by removing edges and merging their vertices until you have the coarsest approximation (like lowest LoD) you could use in the game.

At run time, you load the coarse model and then stream the extra edges and vertices in, progressively, where they're needed - GT6 even animated this process to reduce sudden topology and silhouette "pop-in" as the extra detail was added. The "tessellation" (subdivision) you described was performed on top of that, on the highest detail (original) mesh, and only on certain "loops" of connected edges (edge loops), like the circular rim of a wheel or a headlight bezel etc.

cs-354-project-2-and-compression-20-728.jpg
 
05XR8
It's like, how did we put up with those GT6 models? GTS is really that good.
Someone mentioned that jump from GT2 to GT3. I feel it's still like that. Great work. It's funny, I feel like in playing with my Matchbox cars as a kid in the mid-1970s. Only, this is MORE realistic. :sly:
Click to expand...
Exept the jump from GT2 to GT3 you could have seen a clear upgrade, 3D models were better, less jagged edges, upgraded reflections, a more realistic lighting model; at the time it seemed pretty alien.
With the lifespan from GT4 to GT5 only major upgrade was on the lighting model (introduction of dynamic time and weather) and texture model upgrade (premium cars only).

Now with the PS4's capability this has been further enhanced but you can still see it is limited during gameplay (even with no dynamic time/weather).
With the PS5 I don't expect drastic changes but maybe we might just be able to play with what we see during scapes/ photomode/ gran turismo vista or whatever they want to call it.. in-game with a stable Framerate.
I am 100% sure Dynamic time/ weather will influence the framerate so there will have to be compromises made during in-game either way (let's just hope it is not too much).
 
SDSPOWER01
Exept the jump from GT2 to GT3 you could have seen a clear upgrade, 3D models were better, less jagged edges, upgraded reflections, a more realistic lighting model; at the time it seemed pretty alien.
With the lifespan from GT4 to GT5 only major upgrade was on the lighting model (introduction of dynamic time and weather) and texture model upgrade (premium cars only).

Now with the PS4's capability this has been further enhanced but you can still see it is limited during gameplay (even with no dynamic time/weather).
With the PS5 I don't expect drastic changes but maybe we might just be able to play with what we see during scapes/ photomode/ gran turismo vista or whatever they want to call it.. in-game with a stable Framerate.
I am 100% sure Dynamic time/ weather will influence the framerate so there will have to be compromises made during in-game either way (let's just hope it is not too much).
Click to expand...
No doubt. I had wondered when we would play with the cars as they looked in the GT1 intro. My my my...
 

Similar threads

GTPNewsWire
GT Sport vs GT6 Comparison: The Super Premium Supra
Replies
29
Views
9K
shmogt
shmogt
Groundfish
I would like triple screen support for...
Replies
5
Views
2K
Sunnyb255
S
GTPNewsWire
GT Sport vs GT6: Shelby Cobra 427 Comparison
2 3
Replies
74
Views
12K
IngRobNy
IngRobNy
MarwaanZA0
Deep Appreciation for the Work PD Has Put into GT7
Replies
0
Views
852
MarwaanZA0
MarwaanZA0
MMX
GT Track comparison
Replies
6
Views
3K
MMX
MMX

Latest Posts

Back