How many Polys can the New Consoles Push?
- Thread starter BenQ
- Start date
MechanizedDeath
Regular
Is it too late for me to add to this? 
Number of tris in a mesh is equal to the #verts - 2. You have two naked verts, but you can then form the rest of the mesh by adding one vert for each one. This is actually an easy way to calculating the internal angle of a polygon with N-sides. Take N-2 and multiply times 180 degrees. So a pentagon has 3 * 180 = 540 degrees of internal angles. You see this when you draw a pentagon and using ONE vertice, try to draw a line all the other verts. You'll come up with three tris, or 5-2 = 3. This is how a mesh is formed, and you form solids this way too. I've slept/doodled through my Geometry class this Summer b/c it's basic hs stuff. But dammit if I didn't learn one cool trick during it all. I gotta tell my professor that I actually learned something useful this Summer. PEACE.
EDIT: Oh figs, I'm actually looking at this in 2-dimensions. If you wrap that mesh around to connect the two "naked" verts, you get a 1:1 vert:tri ratio. Take a triangular pyramid for instance. Three verts for the base, and one for the peak. That's 4 verts. How many polygon faces (tris)? You have exactly 4, for a 1:1 ratio. So, there you go. I'd draw something for you, but I don't feel like it. I was feeling so clever too. :?
Number of tris in a mesh is equal to the #verts - 2. You have two naked verts, but you can then form the rest of the mesh by adding one vert for each one. This is actually an easy way to calculating the internal angle of a polygon with N-sides. Take N-2 and multiply times 180 degrees. So a pentagon has 3 * 180 = 540 degrees of internal angles. You see this when you draw a pentagon and using ONE vertice, try to draw a line all the other verts. You'll come up with three tris, or 5-2 = 3. This is how a mesh is formed, and you form solids this way too. I've slept/doodled through my Geometry class this Summer b/c it's basic hs stuff. But dammit if I didn't learn one cool trick during it all. I gotta tell my professor that I actually learned something useful this Summer. PEACE.EDIT: Oh figs, I'm actually looking at this in 2-dimensions. If you wrap that mesh around to connect the two "naked" verts, you get a 1:1 vert:tri ratio. Take a triangular pyramid for instance. Three verts for the base, and one for the peak. That's 4 verts. How many polygon faces (tris)? You have exactly 4, for a 1:1 ratio. So, there you go. I'd draw something for you, but I don't feel like it. I was feeling so clever too. :?
Last edited by a moderator:
Thanks folks. i think i understand better now that I have read Kristof's article... http://www.beyond3d.com/articles/vs/index.php?p=7 what ever happened to old Kristof bTW?
He was my first introduction to Beyond3d when i used to follow the development of the PowerVR based Dreamcast.
He was my first introduction to Beyond3d when i used to follow the development of the PowerVR based Dreamcast.
dukmahsik said:
1280 X 720 is not quite a million pixels per frame ...
thats either 30 million pixels/polys per second on 30 fps game
or 60 million pixels/polys per frame at 60 fps...
so with 440 millon plus polys/pixels to spare you should be able to get VERY high quality polys on screen with great effects...
ShootMyMonkey
Veteran
I can unfortunately say that I've seen worse than 1:3 ratios (by that, I mean repeating the same tris multiple times and using duplicates of degenerate tris which are used to stitch together multiple parts so they go in the same buffer) when artists try to use degenerate tris to stitch together multiple mesh segments. If there's anything I've learned, it's to never doubt the power of artists and/or game testers to do things that you thought would never happen.
Well, it's 921,600 pixels per frame. I'd call that a million.
Mmmm yes... which brings you back to 1Million pixel per frame when you divide the 30fps figure of 30MPixels by 30, or by 60 for the 60fps one....:|
What d'you mean 440M polys/pixels to spare?!
Anyway, next gen consoles seem to have more than enough grunt to be able to push more polygons than they can draw, everyone knows that, much like current gen consoles can calculate a lot more polygons than they can draw - being limited mainly by bandwidth and fillrate. So i'm not sure we can talk about having 440M polys "to spare" so freely.
london-boy said:
Easy L-B! Eat your wheaties man!
if you had a flat 1, 000, 000 dollars in the bank and went to withdraw it... and they gave you $921,600 you wouldnt quite call that a million would you? thus my use of the term.as far as 440 million xenos is said to be able to setup up to 500 million polys max. I was was just illustrating that at most in terms of display xenos would have to output 60 million max.. the rest of that theoretical power (500-60=440) could go towards everything else in the pipeline. Yeah its lowbrow, its laymen but it seems to work conceptually.
blakjedi said:
The rest of the triangle setup power? That's fixed function stuff AFAIK, it's not part of the programmable pipeline I don't think, so savings there would not affect the rest of the chip and what it's doing elsewhere.
Also don't think you'd only ever need no more than 1m polys per frame. Maybe when you map multiple polys to pixels, you'd start questioning how many you need.
overclocked
Veteran
Whats the point of this thread?
