So it is applied for environment to (that needs it) and all characters, cool. Though close tesselation distance for the insane detail it surely will be tweaked/modded. IMO could do with half that detail and 4x the tesselation distance with detail steps. Really what is it 2, 3, 4m polys for that alien?
"Yes, but how many polygons?" An artist blog entry with interesting numbers
- Thread starter Farid
- Start date
brain_stew
Regular
If this finally means the end of jarring LOD transitions then I'll be a very happy man. Its probably the single most jarring artefact of modern game engines if you ask me.
Oh, and I agree with Nebula, the distance and amount of tessellation needs to be tweaked, though its still a very impressive demonstration.
I wonder if they add any form of tessellation to the 360 code? Either way, its nice to see a developer give a PC version of a multiplatform game a significant and worthwile upgrade like this. Just annoyed that its encouraging me to spend money I don't have!
Oh, and I agree with Nebula, the distance and amount of tessellation needs to be tweaked, though its still a very impressive demonstration.
I wonder if they add any form of tessellation to the 360 code? Either way, its nice to see a developer give a PC version of a multiplatform game a significant and worthwile upgrade like this. Just annoyed that its encouraging me to spend money I don't have!
Last edited by a moderator:
That would be awesome but I imagine that it's harder to do tessellation on the 360 in comparisons to a directX 11 card.
Why not there is a specific tessellator unit on 360...so wont that make stuffs easier ?
Infact it'd be rather harder to tessellate on PS3 since tessellation would take up SPE process.
brain_stew
Regular
Well the use of tessellation on the X360 has been far from widespread, there has to be something that's holding back mass (or even minor!) adoption? Perhaps tessellation on the 360 is just too slow or limited to be of any real use? I don't know, the fact that the tessellation units in R600-RV770 aren't supported at all under DX11, is surely evidence that tessellation on the 360 doesn't work in quite the same way?
I'd certainly like some feedback from those in the know as to the reasons why the 360's tessellation unit has seen so little use.
I said nothing about the PS3. I don't know why people would bring it up when there is no reason to. What I meant was that the 360 lacks some of the improvements made it into the tessellator pipeline in AMD's DirectX 10, 10.1 and 11 cards ( Hull Shader and Domain Shader). I'd imagine that these improvements were made to make tessellation easier for devs and improve performance. With the help of memexport devs might be able to make up for the missing improvements but I still haven't been able to get an answer on that.
Last edited by a moderator:
It lacks features the DX11 compliant GPU can do for tesselation. However it might be good to implement it on 360 as they could also do it for the Radeon 3xxx/4xxx series HW tesselator even if the Radeon 3xxx/4xxx tesselation HW is more advanced.
Wow it would appear Valve actually took their time with weapon modeling in L4D2. That pistol looks a head and above better than the one from the first.
Character models have a noticeable bump in quality too, though I don't recall them being all that bad to begin with...? I mean obviously the zombies weren't too detailed up close, that's a given. But I thought Zoe and co had a sizable poly\texture budget allotted to them.
Character models have a noticeable bump in quality too, though I don't recall them being all that bad to begin with...? I mean obviously the zombies weren't too detailed up close, that's a given. But I thought Zoe and co had a sizable poly\texture budget allotted to them.
COPS N RAPPERS
Regular
they always say save the best for last.
MS said that further 360 games are going to progress a whole lot better. (i guess that's why ms made new dev-kits and programs this year)
you know, i never knew 360 had a tessellator.
Last edited by a moderator:
COPS N RAPPERS
Regular
that's amazing.
wow, is this a feature only for PC or will the 360 version have this feature too?
(i notice the "AMD" logo on the bottom left, must be an AMD/ATI expo)
Last edited by a moderator:
Mindblowing.
wow, is this a feature only for PC or will the 360 version have this feature too?
Seems so so far however perhaps they will use it for 360 and ATi 3xxx or newer cards in reduced form. The thing though is that if the tesselation is done by the shaders then it will explain the little use of it for 360 and perhaps no luck then for this game either. But if it is some fixed function hardware then it might. Perfomance priority.
Sebbbi (from RedLynx studio) rised an interesting issue about he use of the 360 on board tesselator:
But there is also the portability problem, it bothering when you paln pc and ps3 renditions.
I fully expect directx11 to help in this regard.
Anyway I fully expect that it has limitation in comparison to what available in newer GPU.Sebbbi answering Granmaster said:
But there is also the portability problem, it bothering when you paln pc and ps3 renditions.
I fully expect directx11 to help in this regard.
Last edited by a moderator:
Actually, I don't think that has to be a problem necessarily. Especially multi-platform titles who generally don't max out SPEs anyway, should be able to use the SPEs for tesselation. So I'm thinking the content creation tools as well as the limitations of directx / other hardware to support on PC are what's holding this back more than anything multi-platform related. I'm increasingly learning that consoles are holding back PC development in the Art department, but not generally in the engine departments - there it seems to be the other way around much more often.
Well tesselation on SPU is clearly possible and even most likely more flexible than what GPU provide but then comes the issue to send that many vertices to the GPU, I don't Know more insights would be welcome but I can clearly see an advantage on keeping stuff on the GPU.
Last edited by a moderator:
Just curious... where did you hear/read this
It is just speculations from my part. Atleast I think so! :smile:
And it really is a shame that Radeon 3xxx/4xxx having tesselation support for all these years without being used. Well I suppose we all should "thank" Nvidia and the TWMBMP program.
There's no clear advantage to keeping things on the GPU on the consoles, I think, at least not by default. It's much more a matter of complexity of the engine in general, and flexibility on the part of the GPU. If we look at the PS3, where the RSX has a weak point in the vertex processing phase, we already see SPE jobs that do a lot of the pre-processing here, detecting which don't have to be drawn, etc. If you look at the whole pipeline, then stuff like animation and such comes even before that. I think it's fairly natural to add tesselation to this part.
I have no idea though how llimited the RSX would be in terms of the rasterization and pixel shading phase, or whether or not the SPEs could do any meaningful work to assist the RSX here also. I would imagine yes, since some games already do this for calculating lighting and reflections, but I don't know how far that can go and where the real-life bottlenecks are. And of course a lot depends also on the game engine - maybe they need the SPEs for something else.
Working with Catmull-Clark subdivs is just a tiny bit different from working with polygons. A lot of the game industry modelers are already familiar with all the workflow issues because they had to learn subdivs for the highres models used in normal mapping several years ago. In fact, a lot of Doom3's content was just simple subdiv models because there was no Zbrush or any other highres sculpting tool. But even Zbrush works with Catmull-Clark, and so do all the other tools we use from modeling through UVs to rendering. It's basically an industry standard by now (and no surprise that it's from Pixar).
The problem is that GPUs don't seem to want to support Catmull-Clark subdivs and implement some rather exotic alternatives instead. Just look at those AvP wireframes... all kinds of strange mesh construction and tesselation.
So the content side's approach is that they'll sit it out until the hardware and engine side gets it 'right'. And the reason for that is simple: movie VFX is using Catmull-Clark subdivs + displacement mapping already and it just doesn't make sense to develop two concurrent high-res artwork pipelines, complete with tools and such.
The problem is that GPUs don't seem to want to support Catmull-Clark subdivs and implement some rather exotic alternatives instead. Just look at those AvP wireframes... all kinds of strange mesh construction and tesselation.
So the content side's approach is that they'll sit it out until the hardware and engine side gets it 'right'. And the reason for that is simple: movie VFX is using Catmull-Clark subdivs + displacement mapping already and it just doesn't make sense to develop two concurrent high-res artwork pipelines, complete with tools and such.
Thanks, that's an interesting inside perspective.
I just looked up Catmull-Clark on Wiki. Maybe for real-time scenarios performance is more critical, and therefore algorithms that more easily scale, and are optically more efficient in terms of taking into account the visible angle and depth? At least, those seem to be difference that I've noticed in the various modern tesselation demonstrations and papers, versus the explanation on wiki, which shows a method that just increases the resolution of the Mesh in general while not taking depth, angle or visibility into account.
I guess what'd we'd need going forward is being able to have content creation tools that properly simulate the rendering of the art taking those parameters into account - maybe being able to input the polygon budget for an object, and then render using the camera's location and distance to provide the parameters to perform the engine's tesselation
I just looked up Catmull-Clark on Wiki. Maybe for real-time scenarios performance is more critical, and therefore algorithms that more easily scale, and are optically more efficient in terms of taking into account the visible angle and depth? At least, those seem to be difference that I've noticed in the various modern tesselation demonstrations and papers, versus the explanation on wiki, which shows a method that just increases the resolution of the Mesh in general while not taking depth, angle or visibility into account.
I guess what'd we'd need going forward is being able to have content creation tools that properly simulate the rendering of the art taking those parameters into account - maybe being able to input the polygon budget for an object, and then render using the camera's location and distance to provide the parameters to perform the engine's tesselation