Crytek Engine Coming To X360/PS3

Tap In said:
how would 360 do this?
Either the hardware tesselator on Xenos, or if that's not full featured in some way use the CPU.
Edit : I'll just explain that by 'not full featured' I mean if the tesselator in Xenos lacks functions found in true 'DX10' class GPUs. I'm not trying to suggest it's not a full tesselator at all, before some six-star starts saying I'm putting down the hardware. It's the first implementation of a hardware tesselator that I know of, and kudos to ATi and MS if they got everything they want in there in the very first attempt! But 'if' it hasn't, the XeCPU can still be used.
 
Last edited by a moderator:
RobertR1 said:
So what happens to all it's DX10/SM4.0 features?

I've heard it claimed that Crytek has said that the only improvements implemented in their DX10 path are optimisations and performance improvements - the effects themselves are visually the exact same. Someone might want to check that, though.
 
Titanio said:
I've heard it claimed that Crytek has said that the only improvements implemented in their DX10 path are optimisations and performance improvements - the effects themselves are visually the exact same. Someone might want to check that, though.

which seems to coincide with Ingenu's post above Re: DX10 in general.
 
Shifty Geezer said:
...

...I mean if the tesselator in Xenos lacks functions found in true 'DX10' class GPUs. I'm not trying to suggest it's not a full tesselator at all, before some six-star starts saying I'm putting down the hardware. It's the first implementation of a hardware tesselator that I know of, and kudos to ATi and MS if they got everything they want in there in the very first attempt!...
shifty

I got this from the "Next_Generation_Graphics_Programming_on_Xbox_360.ppt" that Acert93 pointed to recently (can't find source) but does this tell us anything about whether it's fully featured?

GPU: Hardware Tessellation
3 modes of tessellation

Discrete: integer parameter, triangles only

Continuous: float parameter, new vertices are shifted to their final position

Per-edge: Special index buffer used to provide per-edge factors, used for patches only, cannot support traditional index data as well

Discrete Tessellation Sample
Triangle with tessellation level 4

Continuous Tessellation Sample
Quad with tessellation levels 3, 4, 5

Per-Edge Tessellation Sample
Quad-patch with per-edge tessellation factors of 1, 3, 5, and 7

Also, from dave's Xenos article:
http://www.beyond3d.com/articles/xenos/index.php?p=09

The combination of the shader array and tessellation unit can now make the, oft spoken of but rarely seen, capability of displacement mapping an attainable method to use as this truly becomes a single pass algorithm for Xenos. A simple primitive can be sent to the tessellation unit which is then subdivided into a vertex mesh and then that can be applied to a vertex shader program that does displacement map lookups via the vertex fetch texture units and then the geometry mesh altered according to the sampled values from the texture sampler. Alternatively, if the screen-space projection of the input primitive to the tessellation unit is calculated prior to tessellation then the per-edge tessellation level can be figured out dependant on that projection such that displacement mapping with correct, dynamic level of detail can be achieved.
forgive my ignorance but is this tesselation unit used in current games as part of Xenos normal operation or does it require a DX10 engine, like CryEngine2 to utilize it effectively?
 
Shifty Geezer said:
It's the first implementation of a hardware tesselator that I know of, and kudos to ATi and MS if they got everything they want in there in the very first attempt! But 'if' it hasn't, the XeCPU can still be used.

Just a 6 star (well 5 green :p ) knit-pick: GPU makers have toyed with tesselation in hardware before. D3D10 is a big step in that it seems to be less of a hack and something that makes sense with where the hardware resources have progressed. I guess the question is one of usability and whether it solves problems for developers and gives good results. On the 360 side the question is how robust is it and is it easy to us? Maybe we can hog tie ERP and he could give us some information :LOL:

@ Tap In: It really depends on the game/engine design and how it is implimented, but like most graphical features to be used right you need to plan for it up front. On the PPT side I am not sure of the performance (less I believe triangle setup drop to 250M triangles/s) or the full range of intent of the Xenos tesselation unit, but in some of the early Xenos PPTs from last Spring ATI outlined some uses, namely in conjunction with vertex texturing for single pass displacement mapping. Their example model was a little dinosaur and the displacement map added added "texture" and countour to the creature. From the recent GDC 2006 slides it seems this is still on MS's table in regards to dev evangalism. But seeing as this is not a feature widely used by developers we will probably need to wait for some exclusive games built with Xenos in mind to see such features... which could be never based on the way MS is moving! Or, in the least, until D3D10 development gets underway, in which case if the time span between DX9 API release and games that modestly use a DX9 layer (with fallback; thinking of Far Cry and Half-Life 2 here; Oblivion is the first DX9-only game) it will be a good 24-32 months. Which, interestingly, is the same time frame it takes to make a new game from scratch.

Of course we may find that such features and effects are under powered in regards to the tradeoffs are not good. e.g. the use of tesselation gives a lower visual impact than just dedicating more of the hardware to more traditional approaches.

This is why consoles begin to look really nice in year 3 and 4 and then slow. You spend the first year or 2 just building your first game engine for the hardware, and then it takes a couple years from there to really find that sweet spot of what features work well with the hardware, and which ones are a net loss for your art direction.

Anyhow blabby aside, no one has given an indication of its use yet, but a D3D10 engine is not required--just for developers to code for it. Which may not be until D3D10 games become common place and if the Xenos solution is compatible seeing as it lacks a full blown D3D10 featureset. I guess it would all come down to time and return on investment.
 
Hm....the roundness of the tires in MotoGP 2006 kinda suggest use of the tesselator, or maybe they're actually using a lot of poly's :?:
 
Last edited by a moderator:
Alstrong said:
Hm....the roundness of the tires in MotoGP 2006 kinda suggest use of the tesselator, or maybe they're actually using a lot of poly's :?:

I'd love to find out the used/unsued feature ratio on 360 games released/upcoming so far.

If ms puts the transistors in the hardware and they go unused, they're paying extra basicly for nothing so it's in their best interests they get used but I guess up until this point 360 has had no competition hence no need to truly push these features to make 360 games look "as good" as ps3 games that aren't on the market/trumping games in the works yet.

Can any devs speak to what extent Ms is evangelizing specific featuresets to be used in upcoming games?
 
Good point about the tyres in those MotoGP 2006 screens.
and being reasonably simple geometry, i aimgine it would be a perfect place to use tesselated geometry.

Although somewhere the number of poly's in use on the bikes was mentioned, (i dare say a thread around here), not sure if it was a quite form a dev, if it was a dev quite i would have expected them to make a comment IF they were using tesseleation.

/me goes off to scan the games forum...
 
Alstrong said:
Hm....the roundness of the tires in MotoGP 2006 kinda suggest use of the tesselator, or maybe they're actually using a lot of poly's :?:
$100 on just loads of polygons
all the tires are gonna be the same i believe, ie the same tyre model is used on all bikes (perhaps a couple of variations) thus it makes sense to just store the one highly tesellated model + draw that multiple times in one go, since years with drawing the number of draw calls is more important than the polygon count, eg wether the tyre has 1000 or 10000 polys doesnt really impact performance, since they both will cover roughly the same number of pixels onscreen
 
Back
Top