How many Polys can the New Consoles Push?

[the specs of PS3 and Xbox 360 have not been finalized yet (athough, in Xbox 360 's case, it's specs will have to be finalized very soon)]..

but, as of now;

- Xenos' limit is 500 million triangles/sec. , according to the spec. sheet

- RSX 's limit is 1100 million vertices/sec. , if one extrapolates from G70 's specs., (G70 @ 430 MHz has limit of 860 million vertices/sec. .. and RSX is slated to run @ 550 MHz)
 
Last edited by a moderator:
Wunderchu said:
[the specs of PS3 and Xbox 360 have not been finalized yet (athough, in Xbox 360 's case, it's specs will have to be finalized very soon)]..

but, as of now;

- Xenos' limit is 500 million triangles/sec. , according to the spec. sheet

- RSX 's limit is 1100 million vertices/sec. , if one extrapolates from G70 's specs., (G70 @ 430 MHz has limit of 860 million vertices/sec. .. and RSX is slated to run @ 550 MHz)

whats the difference between triangles/sec and vertices/sec
 
hasanahmad said:
whats the difference between triangles/sec and vertices/sec

AFAIK, for these purposes, they are the same. If you "push" x number of vertices per sec you can pretty much claim to push the same x number of triangles. 1 triangle can = 1 vertex, effectively.

I'm confused though, if G70's figures are the transform rate or setup limit.
 
Titanio said:
AFAIK, for these purposes, they are the same. If you "push" x number of vertices per sec you can pretty much claim to push the same x number of triangles. 1 triangle can = 1 vertex, effectively.

I'm confused though, if G70's figures are the transform rate or setup limit.

Excuse me, but 1 triangle does NOT = 1 vertice.

A triangle is comprised of 3 vertices, but I'm not implying that you should divide the PS3's spec by 3, as you can have multiple triangles sharing one vertice. But you can NOT claim that "1 triangle can = 1 vertex, effectively."
 
lets get the grammar right here please:


Triangle: The plane figure formed by connecting three points not in a straight line by straight line segments; a three-sided polygon.

Vertex: The point at which the sides of an angle intersect.


1 Triangle has 3 vertices , top, bottom left and bottom right

Now keep discussion going
 
BenQ said:
But you can NOT claim that "1 triangle can = 1 vertex, effectively."
sure you can..

take, for example a 3 sided pyramid, with a triangular base, like in this pic.:
tetrahedron.jpg

[source: http://www.jaist.ac.jp/~g-kampis/Lecture_One ]

there are 4 vertices, & 4 triangles in this pic.
 
When creating a mesh, you can add a triangle by adding just 1 vertex. Think of a ribbon. You start with the first triangle made of 3 vertices. The ratio triangles:vertices = 1:3. The next triangle uses two already existing verticies and adds one more. The ratio's now 2:4. Same with the next triangle, giving 3:5. Continue this on and on, and the ratio triangles:vertices gets closer and closer to 1:1. It has been said on this forum by others that you can even get a ratio of less than 1 vertex per triangle. Someone posted the average triangle:vertex ratio for a game too, which was between 1 and 2 if memeory serves. But in peak rates it's fair to consider 1 vertex=1 triangle as that's an ideal situation for pushing polygons.
 
BenQ said:
That's not fair, unless the only thing your planning on rendering is pyramids. You can't claim a 1:1 ratio with anything more complex than a pyramid.
heh ... I guess my example does not really apply too well to what we are discussing .. I apologize :D



Shifty Geezer wrote a more relevant explanation :)
 
Shifty Geezer said:
When creating a mesh, you can add a triangle by adding just 1 vertex. Think of a ribbon. You start with the first triangle made of 3 vertices. The ratio triangles:vertices = 1:3. The next triangle uses two already existing verticies and adds one more. The ratio's now 2:4. Same with the next triangle, giving 3:5. Continue this on and on, and the ratio triangles:vertices gets closer and closer to 1:1. It has been said on this forum by others that you can even get a ratio of less than 1 vertex per triangle. Someone posted the average triangle:vertex ratio for a game too, which was between 1 and 2 if memeory serves. But in peak rates it's fair to consider 1 vertex=1 triangle as that's an ideal situation for pushing polygons.

I don't agree. With a single triangle, each one of the vertices is made up of the intersection of 2 angles. Now slap another trianle beside the first. The second triangle can share 2 of the vertices, but now those 2 vertices are comprised of the intersenction of 3 angles rather than 2.

As far as system perforance goes, is the complexity of the vertice irrelevant? Does having a vertice comprised of 3,4,5,ect angles NOT cost any more than a 2 angle vertice?
 
BenQ said:
That's not fair, unless the only thing your planning on rendering is pyramids. You can't claim a 1:1 ratio with anything more complex than a pyramid.

Yes, you can. Or virtually so. As Shifty points out, if constructing a mesh, you can add new triangles by simply adding one vertex. Every triangle has 3 vertices, but in a mesh they are rarely unique, non-shared vertices. The more triangles you have sharing vertices, the closer you get to that 1:1 ratio.
 
hasanahmad said:
that shape is of a quadrilateral, not a triangle
if one takes the dotted line as behind the solid ones, the pic. can be seen as a 3D representation of a triangular prism with a triangular base (AKA: a tetrahedron)

but, yes, it can also be seen as a quadrilateral :D
 
BenQ said:
That's not fair, unless the only thing your planning on rendering is pyramids. You can't claim a 1:1 ratio with anything more complex than a pyramid.

Sure you can.

The way the GPU works, as long as you don't need to make any state changes, you can pretty much draw as many triangles you want in your triangle strip. This is how you usually make polygon meshes like say this:

sphere3d.gif
 
Last edited by a moderator:
heres a nice comparison:

Transistor counts: 6800 had 222m, X850 had 166m, 7800 as 302m
Manufacturing Process: 6800 and X850 used 0.13, 7800 uses 0.11
Clockspeed: 6800 was 425 Mhz, X850 was 540mhz, 7800 is 430mhz
Number of pixel pipes is 6800 and X850 both had 16, 7800 has 24
Number of Texturing units is 6800 and X850 both had 16, 7800 has 24
Number of vertex pipelines is 6800 and X850 had 6, 7800 has 8
Peak Pixel Fill rate performance is 6800 had 6.8, X850 had 8.6 and 7800 has 6.9 Gpix/s
Peak Texture Fill rate performance is 6800 had 6.6, X850 had 8.6 and 7800 is 10.32 Gpix/s
All 3 memory interfaces are 256 bit
Memory clock speed is 6800 had 1.1Ghz, X850 had 1.19, 7800 has 1.2
Peak memory bandwidth is 6800 had 35.2, X850 had 37.8 and 7800 has 38.4 GB/s
 
Titanio said:
Yes, you can. Or virtually so. As Shifty points out, if constructing a mesh, you can add new triangles by simply adding one vertex. Every triangle has 3 vertices, but in a mesh they are rarely unique, non-shared vertices. The more triangles you have sharing vertices, the closer you get to that 1:1 ratio.

That is true, but is it fair to simply not account for the increased complexity of those vertices?

Something smells fishy to me.

For every other console in the past, they release how manys poly per second it can push in the spec sheet, this is the first time I have seen vertices/second rather than polys/second. Which leads me to wonder why.

The answer I come up with is that M$ released their numbers first and then Sony's PR team decided using vertices/second rather than polys ( and most people don't even know the difference ) looks much better on paper.

I want to see how many polys the PS3 an push OR how many vertices the Xbox 360 can push. Comparing polys to vertices seems designed purely to misslead.
 
Wunderchu said:
if one takes the dotted line as behind the solid ones, the pic. can be seen as a 3D representation of a triangular prism with a triangular base (AKA: a tetrahedron)

but, yes, it can also be seen as a quadrilateral :D

yes but there is an important difference: triangle has 3 sides, quadrilateral has 4 sides. If you take that into account, the numbers end up same in terms of efficiency as well as triangles vs quadrilaterals
 
hasanahmad said:
yes but there is an important difference: triangle has 3 sides, quadrilateral has 4 sides. If you take that into account, the numbers end up same in terms of efficiency as well as triangles vs quadrilaterals
if one takes that pic. I posted as a 3D shape, then there are 4 triangles there



actually, come to think of it, there are 4 triangles, if one takes it as a 2D shape too.. heh, I didn't notice at first
 
Last edited by a moderator:
BenQ said:
That is true, but is it fair to simply not account for the increased complexity of those vertices?

Something smells fishy to me.

For every other console in the past, they release how manys poly per second it can push in the spec sheet, this is the first time I have seen vertices/second rather than polys/second. Which leads me to wonder why.

The answer I come up with is that M$ released their numbers first and then Sony's PR team decided using vertices/second rather than polys ( and most people don't even know the difference ) looks much better on paper.

I want to see how many polys the PS3 an push OR how many vertices the Xbox 360 can push. Comparing polys to vertices seems designed purely to misslead.

AFAIK, that 1.1B number isn't from Sony, rather it is an extrapolation. The G70 is capable of setting up 2 triangles per clock or 860 million. Ramp the clockrate up to 550mhz and you get 1.1B

That is my understanding, atleast.

I wouldn't really consider setup rates to be a metric of "power", but that is just me.
 
Back
Top