Intresting P10 next-gen info...

[alexsok]

http://www.xbitlabs.com/news/story.html?id=1033172802

Here's what caught my attention:

Sounds intresting...

 

[megadrive0088]

yeah its interesting. thanks for posting.

As I've said, Creative/3DLabs might come out with a DX9 capable part next year. A P15 or P20 if you will. P10 is sub R300/Nv30 performance
as well as in features. wouldnt make much sense to have a consumer P10 based card out now. at best, it would rival Parhelia. I would have liked to have seen a consumer P10 but I guess its not to be.

 

[PSarge]

Hmmm.....interesting.

I wonder where that slide came from.

 

[Kristof]

Anyone want to try and explain nano-threading to me ?

K-

 

[CMKRNL]

My understanding is that nano-threading refers to a simultaneous multi-threading implementation. This allows them to switch to a different execution stream when the engine is stalled on a memory fetch.

 

[Dave Baumann]

Also of note is the first point: "Combine programmable arrays for more programming generalityâ€￾ -- sounds like something 'SA' brought up a while ago; rather than a vertex shader and pixel shader, how about a general purpose shader array?

[edit] Here's the thread, in which SA said:

- The merging of vertex shaders and pixel shaders into a single, completely programmable, high-level, floating point, 3D hardware language. Some vendors are likely to merge the floating point hardware functionality of pixel processors and vertex processors as well. In this case they would likely allocate the merged set of floating point processors dynamically much like P10 allocates them dynamically for just vertex shaders. --- timeframe: late 2003 - 2004

 

[Hyp-X]

We'll have dynamic branching in the vertex and pixel pipeline (P10 has that already).
We'll have a floating-point pixel pipeline (R300 has that already).
The vertex pipeline will be able to fetch textures (it's in DX9 which means its coming soon).

Once all of this is done, there's no more reason to separate pixel and vertex pipelines - it's almost completely the same.

In current GPUs (or VPUs) the pipelene has vertex processing and pixel processing both with its dedicated execution units.
In most cases rendering is either fillrate limited or vertex calculation limited, which means one of the two units is not used to 100% of potentials.
Sharing the hardware beetween vs/ps means better utilization of the hardware.

 

[alexsok]

We'll have dynamic branching in the vertex and pixel pipeline (P10 has that already).

You sure about that? I haven't been able to get a good answer on that...
If that's indeed the case, i'm very impressed by what 3DLabs were able to accomplish with P10! I mean, R300 doesn't support any kind of branching in PS, nor does NV30 (although it does support condition codes, with the help of which, many things that can be accomplished with branching could also be accomplished with them).

 

[JF_Aidan_Pryde]

As Democoder said before, SA probably meant the combined pixel / vertex shading *language* rather than physical units.

 

[Dave Baumann]

Some vendors are likely to merge the floating point hardware functionality of pixel processors and vertex processors as well.

 

[nggalai]

Anyone want to try and explain nano-threading to me ?

K-

Easy--"add meaningless marketing blurb here"



Seriously, now--for a technology overview, this slide is as usable as a glass hammer for carpeting work . . . where does it come from, exactly? Is it an x-bit exclusive, or something (semi-)official from Creative / 3DLabs?

ta,
-Sascha.rb

 

[Humus]

Some vendors are likely to merge the floating point hardware functionality of pixel processors and vertex processors as well.

Some vendors are likely to merge the floating point hardware functionality of pixel processors and vertex processors as well.

In other words, not neccesarily the physical units, but probably a unified language at least.