"Yes, but how many polygons?" An artist blog entry with interesting numbers

But the same unknown sense that allows people to differentiate between real and CGI humans will also make them see the difference between an SH3 character and a CGI character. I thought this is obvious.

Asking friends doesn't count.
 
67K is a loot, keep in mind it has to be skinned, used for player shadows and such things as well.

Well skinning is one thing, but I don't see why they couldn't use a lower LOD model for the shadows. Don't many games do this anyway?
 
Well skinning is one thing, but I don't see why they couldn't use a lower LOD model for the shadows. Don't many games do this anyway?
If you don't care about breaking self shadowing you can do it
 
But the same unknown sense that allows people to differentiate between real and CGI humans will also make them see the difference between an SH3 character and a CGI character. I thought this is obvious.

Asking friends doesn't count.

What about me? And why dont my friends or other people who seem to feel the same way count?

edit: I also disagree that the same can be applied as much on CGI vs Real Time as it can in real life vs CGI.

Perhaps to you since you are working in the industry yes.


Also depends on what kind of CGI you are referring to and compare.
 
Last edited by a moderator:
What about me? And why dont my friends or other people who seem to feel the same way count?

Apparently, when Laa-Yosh says "average person," neither you nor anyone else you know actually counts. I didn't think SH3 was done in real-time, either.
 
Shame we don't have any Melbourne House devs here.

I would love to know how many polys Transformers for ps2 pushed.

I've got contacts some of them so I'll see if I can track some figured down in the next few days. ;) I wouldn't expect them to be too high since it did rely very heavily on LOD to get the massive scenes but we didn't care because they where still awesome.
 
I'll probably get around to proper DOA3/DOAU comparisons over the weekend. DOAX just had 8 characters and far fewer models (not including swim suits though! :oops: ), so here goes...

DOAX showed different models for certain hat/hair configurations. With hat accessories they removed poly's that wouldn't be visible. As far as I could tell, the characters use ~30-35 textures (hair, face, eyes, body, mouth interior, and a bunch of things I can't even identify because they just look like some sort of flesh). Non-alpha textures use DXT1 compression and the hair ones use DXT5. Dimensions range from 64x128 to 256x512.

Where I say hat, I mean that the model has been adjusted to accomodate a hat.

Code:
Tina00    13765    normal
Tina01    13753    hat
       
Kasumi00    15015    long hair       
Kasumi01    14761    long hair hat
Kasumi02    15463    braided      
Kasumi03    15269    pony tail      
Kasumi04    15341    braided hat      
Kasumi05    15147    pony tail hat 
Kasumi06    15151    bow 
Kasumi07    15029    bow hat
       
Helena00    14515    normal
Helena01    13685    hat
Helena02    13572    small bow & hat       
               
Lei Fang00    16988    braid
Lei Fang01    16802    braid hat
Lei Fang02    14098    pony tail up
Lei Fang03    13956    pony tail up hat       
Lei Fang04    13954    pony tail down
Lei Fang05    14236    hair down
Lei Fang06    14090    hair down hat
       
Ayane00    14182    normal (bandana)       
Ayane01    13622    hat
Ayane02    13929    no hat
       
Hitomi00    14721    normal 
Hitomi01    14523    hat
Hitomi02    14789    normal + pink um.. thing on her head

Christie00    14366    normal
Christie01    14108    hat
       
Lisa00    15301    normal
The poly counts could be slightly off from extraction and importing though.

So it seems that Team Ninja was probably figuring about 20K polys per character for worst case scenario including the swim suits and accessories (shoes, head/hair accessory, wrist accessory). I'll see about the swimsuits later. Skimpier suits for higher performance!

Oh and the volleyballs use mostly the same model, ~815 poly's. A couple are different for some reason (864 Gold Ball, 879 Silver Ball), appearing to have a lower poly structure within it (for physics or collision detection?). The Silver Ball is also a tad smaller than all the other ones...

The poly counts in DOAU should be more interesting. Stay tuned...
 
nAo said:
If you don't care about breaking self shadowing you can do it
You can still get away with it if you're careful with how you construct the conservative shadow model. That said I can see how it would be much harder to do that with animated models.
Although on the other hand - most modern games have shadow resolution artifacts in very nasty contrast to fine normal mapped details that ruin the self-shadowing look anyhow(at least in closeups) - it's similarly ugly to volume shadow artifacts with goraud interpolated shading, except worse.
 
I'm actually quite suprised on the numbers!
like jak and daxter 2...
i never thought they were pushing so many!
I'm intrigued on how many polygons were used on Zone of the Enders 2 and Rogue Galaxy!
 
Hi Stephan !

Ok, thanks. But can you make sure Diamant sees my post, something like pm him. I would appreciate ( if he can ) having an answer in my email.

Thank you very much and excuse me.
 
He has already answered you ;) Check the other thread.

EDIT:
DIRECT LINK but the other link above would have worked as well, if you'd have looked in the right upper corner ;)
 
I was wondering about that myself (this morning!), but I haven't had a look at the disc contents yet. I'll have to check later in the week though. (I didn't even have time this past weekend to get some counts from DOAU/DOA3 . :oops: )
 
Back
Top