What improvement would you most like to see, in Xenon?

If the leaked specs for Xenon are true, what improvement would you most like to see?

  • higher bandwidth EDRAM (currently "32 GB/sec")

    Votes: 0 0.0%
  • more main RAM (currently "256+ MB")

    Votes: 0 0.0%
  • higher bandwidth main RAM (currently "22.4+ GB/sec")

    Votes: 0 0.0%
  • higher VPU pixel fill-rate [currently "4+ gigapixels/sec (16 gigasamples/sec with 4× AA)"]

    Votes: 0 0.0%
  • higher VPU vertex rate (currently "500+ million vertices/sec")

    Votes: 0 0.0%
  • higher VPU geometry-rate (currently "500+ million triangles/sec")

    Votes: 0 0.0%
  • greater shader processing capabilities (currently "48 billion shader operations per second")

    Votes: 0 0.0%
  • faster game disc drive (currently "12× DVD")

    Votes: 0 0.0%
  • higher capacity MUs (currently"64 MB")

    Votes: 0 0.0%

  • Total voters
    191
More RAM, more RAM, more RAM.

Faster RAM would be nice too, but not as critical.

The EDRAM is plenty fast for what its intended to do.

More EDRAM would be nice, but you can get around that though various means.

Higher pixel fillrate would be useful, but not as useful when the GPU's designed to go all out on shaders instead. As long as you have enough.
 
bleon said:
Hard drive in every console without having to pay extra

It just means you pay extra when you buy the console instead. In what way is that better? :D

Anyway, I wanted to vote "smaller case size", but that option wasn't listed. ;)
 
It just means you pay extra when you buy the console instead. In what way is that better?

I always thought of the HD and netcard in Xbox as free gifts from MS. MS offers more power for the same price as Sonys PS2 as an incentive to buy. If Xenon ends up being the least powerful console with HD and backwards compatibility left out then that makes Xenon less attractive.
 
More Ram would give ANY console the biggest immediate bump. So i opted for that, seen how everything else is pretty much abstract.

One thing, how can the system process 500M vertices per sec, and 500M polygons per sec too? Is that possible? Aren't the 2 numbers supposed to always be different, since a polygon is made of 3 vertices at least (2 with strips)?
 
london-boy said:
One thing, how can the system process 500M vertices per sec, and 500M polygons per sec too? Is that possible? Aren't the 2 numbers supposed to always be different, since a polygon is made of 3 vertices at least (2 with strips)?

Indexed vertex buffers.

http://msdn.microsoft.com/library/e...enderingfromvertexindexbuffers.asp?frame=true

In any mesh, triangles share vertices. If you order your vertices right, the GPU can use the post-transform cache to avoid transforming the vertex again.

So each 1 additional vertex = 1 additional triangle.

tstrip2.gif
 
aaaaa00 said:
london-boy said:
One thing, how can the system process 500M vertices per sec, and 500M polygons per sec too? Is that possible? Aren't the 2 numbers supposed to always be different, since a polygon is made of 3 vertices at least (2 with strips)?

Indexed vertex buffers.

http://msdn.microsoft.com/library/e...enderingfromvertexindexbuffers.asp?frame=true

In any mesh, triangles share vertices. If you order your vertices right, the GPU can use the post-transform cache to avoid transforming the vertex again.

So each 1 additional vertex = 1 additional triangle.

tstrip2.gif

I knew that, but u can't always use that, and in the end the number of vertices is still higher than that of the polys.
 
london-boy said:
I knew that, but u can't always use that, and in the end the number of vertices is still higher than that of the polys.

Correct, however it is possible to render 500 million trianges with 500 million vertices, and that's why the spec is quoted that way.

Anywhere you see someone mention # of triangles per second for any 3D hardware, what they probably mean in reality is # of vertices per second.
 
aaaaa00 said:
[quote="london-boy]I knew that, but u can't always use that, and in the end the number of vertices is still higher than that of the polys.

Correct, however it is possible, and that's why they say it can be done that way.[/quote]

Theoretically.... But even then, unless your 3D world has each of its objects and characters attached to each other, it's just impossible...
 
london-boy said:
But even then, unless your 3D world has each of its objects and characters attached to each other, it's just impossible...

If each of your seperate meshes has 5000 triangles in it, then the extra 2 vertices you transform per mesh doesn't make a whole lot of difference.

Yes, in reality it won't be 100%.
 
aaaaa00 said:
london-boy said:
But even then, unless your 3D world has each of its objects and characters attached to each other, it's just impossible...

If each of your seperate meshes has 5000 triangles in it, then the extra 2 vertices you transform per mesh doesn't make a whole lot of difference.

Yes, in reality it won't be 100%.


The difference would change depending on the situation yeah. But take a scene with thousands of different objects and characters on screen, all of which detached from each other and with it's own set of physics (which is what i'd expect next gen) and there you go...
 
london-boy said:
The difference would change depending on the situation yeah. But take a scene with thousands of different objects and characters on screen, all of which detached from each other and with it's own set of physics (which is what i'd expect next gen) and there you go...

It's no different for any other hardware. People tend to quote triangles when they mean verticies. People also tend to quote polygon when they mean triangle. :)

If I have 1000 meshes in my scene, each with 4998 triangles (5000 vertices), I will have 5 million vertices in my scene, and 4,998,000 triangles. That's a 0.1% difference.

Here's a good powerpoint that will explain everything:

http://research.microsoft.com/~hoppe/tvc.ppt

If you do your optimization right, you can get > 1 triangle per each additional vertex.
 
Back
Top