Nick Burton (RARE) talks about the tech under Kameo

[expletive]

Hmmm, I've a bit more malice than you. I've a feeling that either it's tougher to dev. for or something else. Kameo runs at 30fps, iirc, and most models are seriously lacking in the poly dept(note: with regards to nextgen expectations), and it's in things like the overall roundness, I think even in cutscenes, which should not be that hard to model(using that subdivision gphx technique and some tweaking, though I'm not familiar with it or its use, so I dunnoh.). Square-enix, as stated managed to get a 60fps realtime cutscene out of cg source models in about two months(meaning the models obviously took less than that to adapt for realtime use), with just a couple guys doing a simple demo, not a full blown large dev. team.

edit:
Remember it's also not only final, but betas, and they probably were among the first to receive betas(being a first party dev.). They also had access to intimate spec estimates/details, and s/w emus and whatnot even prior to that.

I think you start with something like "based on what we know, we can safely push 100,000 polys per scene". Then you get the final hardware, youve got 2.5 months with it to do EVERYTHING you need to when you get final hardware. You determine with the final kit that you could push 400,000 polys per scene. Now it would be easy to say "increase all the models by a factor of 4" but i really dont think its that simple with such a compressed timframe.

I personally dont see the point of the square enix comparison. Theres so many differences in the two scenarios - you may as well be comparing what any dev team on any hardware at any point in time did in 2 months.

 

[RedBlackDevil]

You determine with the final kit that you could push 400,000 polys per scene.

at 60 fps it can output at max 8 milions per scene

the final kit is the only with xenos, I think that the devs need 6-10 months with this in his hands to get the most

so a title that came from xbox and that came before from cube, and that was completed and published 2 months after the final kit arrives, can't be rappresentative

but 12000 models at time is awesome, and they put normal+offset maps in all the materials

Kudos to rare

Gameplay can be discussed, but this is not the right board section, just move to the "console games" then, and don't derail the thread

 

[Shifty Geezer]

Now it would be easy to say "increase all the models by a factor of 4" but i really dont think its that simple with such a compressed timframe.

We also don't know that 4x being drawn means the same demands as 4x the poly count. It could mean that each NPC was being drawn 4x over itself, but being instanced added negligable demands to all but fillrate, of which XB360 has great buckets. Without a breakdown of poly counts, shader instructions and so forth, there's no way of knowing what the limiting factor is. eg. On Xenos more complex pixel shaders takes away from available vertex processing. Perhaps the game was diverting resources elsewhere? Also if it's running 30 fps, they surely can't be tapping only a quarter of the power. Perhaps it's CPU limited? Or somewhere on the graphics pipeline limited? Whatever, I don't see that an accidental 4x drawing can be equated to 4x the poly budget.

Personally I think Kameo looks very good (in screenshots. I've not seen in running for real). There's quite a lot of background scenery in some scenes which can quickly consume poly-budgets, and if you have to budget for those big scenes, you won't have as much to use on main characters which shows on simpler scenes. I certainly see it as a next-gen 'platformer' well beyond current gen and a very good launch title regards visual appeal.

 

[RedBlackDevil]

We also don't know that 4x being drawn means the same demands as 4x the poly count. It could mean that each NPC was being drawn 4x over itself, but being instanced added negligable demands to all but fillrate, of which XB360 has great buckets.

not only the 12K overdrawed models are instanced, but the 3K in game are too
it's clear that they uses instancing, without this, the whole thing is impossible to obtain

the good is that, as far as I know, instanced models can be not the same thanks to "texture array application" feature [pag. 7 of ED05-Xenos ATI doc]
in istancing we can have different models using unique seeds, and each instanced group can use different "texture arrays" and different displaced subdivision surfaces via tessellator

 

[RedBlackDevil]

in istancing we can have different models using unique seeds, and each instanced group can use different "texture arrays" r

 

[overclocked]

Maybe they will take some of the design that apparently was thossed away and release it at live, would be a great idea imo.

 

[CountZeroInt]

Both this and PD0 were Xbox 1 games and it shows, Viva Pinata is only one of Rare's next games...I'm sure whatever else they are making looks even better.