This is a split board - You can return to the Split List for other boards.

I'm starting to see more and more how horrible & outdated 3d model animation

#11SinisterSlayPosted 8/19/2014 2:17:17 PM
CC Ricers posted...
SinisterSlay posted...
JKatarn posted...
It turns out that programming dynamic animation and character interaction with dynamic 3D environments (while not chewing through CPU/GPU resources cycles) is not exactly a trivial task. Ideally we'd never see any clipping/animation errors, but these are problems that have existed since the advent of 3D consoles/3D accelerator cards. If you have some revolutionary collision/animation model I'm sure the top studios would be glad to hear of it.


Yeah at the moment collision detection uses hit boxes instead of the actual faces of the polygons because that requires less calculations.


Sometimes there is a middle ground between both types of collision detection, where it first does a "broad" scan of what hitboxes touched something, and then check the polygons in each of those hitboxes to see if one of them touched something. A collision with the entire model is found with the first polygon collision it finds. The worst case is that it has to check every box and every polygon of the model, but that will not happen in most frames.


I wonder how X3 TC works with those giant stations that if you find certain spots, can actually fly inside them (or between panels)
---
He who stumbles around in darkness with a stick is blind. But he who... sticks out in darkness... is... fluorescent! - Brother Silence
#12ThisGuy101Posted 8/19/2014 2:24:04 PM
jamieyello3 posted...
Yeah of course I'm not talking about full polygon detection, that's ridiculous. I'm talking about what JKatarn said with the dynamic animation. That should really be a standard thing. So if a head is on the ground, there's really no excuse for it to just pop into it's dieing animation from standing up, which you're going to see in most any game.

What would be nice would be if it could be completely physics based, like Sumotori Dreams except perfected on and standard across the industry.


What you're thinking of is Rockstar and its Euphoria engine.
---
AMD FX-6300 | Gigabyte GA-990FXA-UD3 ATX AM3+ | Kingston HyperX 8GB DDR3-1600 | MSI Radeon R9 270X 2GB TWIN FROZR
My first gaming rig
#13KillerTrufflePosted 8/19/2014 2:32:43 PM
SinisterSlay posted...
JKatarn posted...
Also tc: http://en.wikipedia.org/wiki/Collision_detection if you read through that you'll see it's not exactly a simple problem to solve. Doing per face/polygon collision checks on everything in a scene would likely bring everything but a government supercomputer to a grinding halt (depending on scene complexity of course).


Probably some of those Nasa computers could do it.


Even accounting for tessellation? I can just imagine the processing power needed to do per-face collision detection on even a single barricade in Crysis 2.

http://techreport.com/r.x/crysis2/barrier-dx11-full.jpg

http://techreport.com/r.x/crysis2/barrier-dx11-mesh.jpg
http://techreport.com/r.x/crysis2/barrier-side-mesh-620.jpg
http://techreport.com/r.x/crysis2/newbarrier-mesh.jpg

http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/2

Those polys. >_> It would take an insane amount of processing power to calculate per-face collision on however many billions of polys that is. And that's just one single, simple barricade.
---
"How do I get rid of a Trojan Horse?" -Sailor_Kakashi
"Leave it outside the gates of Troy overnight." -Davel23
#14JKatarnPosted 8/19/2014 4:58:52 PM
jamieyello3 posted...
Yeah of course I'm not talking about full polygon detection, that's ridiculous. I'm talking about what JKatarn said with the dynamic animation. That should really be a standard thing. So if a head is on the ground, there's really no excuse for it to just pop into it's dieing animation from standing up, which you're going to see in most any game.

What would be nice would be if it could be completely physics based, like Sumotori Dreams except perfected on and standard across the industry.


The reason every object in every game isn't physics based is because it would take a LOT of CPU power to do that for every object (depending on the complexity of the calculations of course). That's why you typically see computationally "cheaper" ragdoll physics systems that are easier on the CPU (and glitchier).
---
Asus P8Z68-V LE | Core i7 2600K | 8GB G.Skill Ripjaws DDR3 | Gigabyte GeForce GTX 660 Windforce OC
PS3 | PS2 | PSP| Wii | 3DS | DS | X-Box 360 | X-Box | NES
#15LordSeiferPosted 8/19/2014 5:00:00 PM
its time we move to 4d
---
^ this
#16SinisterSlayPosted 8/19/2014 9:33:34 PM
KillerTruffle posted...
SinisterSlay posted...
JKatarn posted...
Also tc: http://en.wikipedia.org/wiki/Collision_detection if you read through that you'll see it's not exactly a simple problem to solve. Doing per face/polygon collision checks on everything in a scene would likely bring everything but a government supercomputer to a grinding halt (depending on scene complexity of course).


Probably some of those Nasa computers could do it.


Even accounting for tessellation? I can just imagine the processing power needed to do per-face collision detection on even a single barricade in Crysis 2.

http://techreport.com/r.x/crysis2/barrier-dx11-full.jpg

http://techreport.com/r.x/crysis2/barrier-dx11-mesh.jpg
http://techreport.com/r.x/crysis2/barrier-side-mesh-620.jpg
http://techreport.com/r.x/crysis2/newbarrier-mesh.jpg

http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/2

Those polys. >_> It would take an insane amount of processing power to calculate per-face collision on however many billions of polys that is. And that's just one single, simple barricade.


Isn't tesselation the gpu guessing where to add more 3d faces based on the colors in the texture?
That would be impossible because the game doesn't truly know every location that has tesselation.
I'm not sure if we will ever be able to do that precise.
---
He who stumbles around in darkness with a stick is blind. But he who... sticks out in darkness... is... fluorescent! - Brother Silence
#17JKatarnPosted 8/20/2014 2:54:45 PM
KillerTruffle posted...
SinisterSlay posted...
JKatarn posted...
Also tc: http://en.wikipedia.org/wiki/Collision_detection if you read through that you'll see it's not exactly a simple problem to solve. Doing per face/polygon collision checks on everything in a scene would likely bring everything but a government supercomputer to a grinding halt (depending on scene complexity of course).


Probably some of those Nasa computers could do it.


Even accounting for tessellation? I can just imagine the processing power needed to do per-face collision detection on even a single barricade in Crysis 2.

http://techreport.com/r.x/crysis2/barrier-dx11-full.jpg

http://techreport.com/r.x/crysis2/barrier-dx11-mesh.jpg
http://techreport.com/r.x/crysis2/barrier-side-mesh-620.jpg
http://techreport.com/r.x/crysis2/newbarrier-mesh.jpg

http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/2

Those polys. >_> It would take an insane amount of processing power to calculate per-face collision on however many billions of polys that is. And that's just one single, simple barricade.


Wow....that seems like such a waste of fillrate for a relatively simple (geometrically speaking) object....why would they use so many polys vs. a simple mesh for each side with a more advanced/parallax map texture is beyond me.
---
Asus P8Z68-V LE | Core i7 2600K | 8GB G.Skill Ripjaws DDR3 | Gigabyte GeForce GTX 660 Windforce OC
PS3 | PS2 | PSP| Wii | 3DS | DS | X-Box 360 | X-Box | NES
#18KillerTrufflePosted 8/20/2014 3:21:28 PM
SinisterSlay posted...
Isn't tesselation the gpu guessing where to add more 3d faces based on the colors in the texture?
That would be impossible because the game doesn't truly know every location that has tesselation.
I'm not sure if we will ever be able to do that precise.


Yeah, it's all done on the GPU to add detail that isn't technically "there" to simpler models. However, physics can be calculated on the GPU too, so it's totally within the realm of possibility to factor tessellated surfaces into physics calculation. The biggest block to that is processing power. It would take a *crapload*, or else a totally rewritten method. But I totally see it as possible.
---
"How do I get rid of a Trojan Horse?" -Sailor_Kakashi
"Leave it outside the gates of Troy overnight." -Davel23
#19NingishzidaPosted 8/20/2014 6:20:34 PM(edited)
its pretty damn absurd to model something as boring as a barricade with so many polys. these people need to use ingenuity to make textures give the impression of more polys. that takes thought, though. and talent.

but then, its all about selling GPUs - this is all about stimulating hardware sales more than the quest for photo-realism, (and of course photo-realism more than game-play in the wider sphere).