Xbox 2 to have 4.0 Shaders?

[Megadrive1988]

http://www.beyond3d.com/forum/viewtopic.php?t=8959&postdays=0&postorder=asc&start=0


DirectX10 / DirectX Next:

Goals
-Customers
-I/O Model
-Shader Programming Model
-Surface Models
-Topology Operations
-Frame-Buffer access in pixel shader
-Virtual Video Memory (yay!)

Goals
-Enable scenarios currently blocked
+Many desirable functions must still be performed on the CPU
-Add generality to reduce number of future scenarios that may get blocked
-Optimize interface to future OSs
+Better primitive throughput performance
+Better state change performance
+Better interoperation of multiple apps/processes

DirectX Graphics Trends
-DirectX Next: dynamic geometry/topology modification
+Move CPU/GPU transition to point of very low bandwidth
+Enable new capabilities for new GPU uses

General Programming Model
-Integer instruction set
-Support for more programming constructs
+stacks and arrays
-More flexible memory addressing
+less texture specific
-Resources are unlimited
+temporary memory store
+instruction count
+iterators, streams, etc.
-Non-real-time usage can be slow
+Performance may drop drastically below 640x480 10 Hz
-Improves hardware as an HLSL target
+All algorithms will at least compile instead of hitting instruction count limits
-Identical for all shader model 4.0 shaders


Since DX10 seems to have 4.0 Shaders, and R500 is said to be a DX10 part, and XBox 2 should have a derivative of R500, logic would say that Xbox 2 will have 4.0 shaders 8)


*waits for someone to blow a hole through that chain of assumptions*

 

[cthellis42]

Seems fine to me, though I'm really starting to wonder about the delay between R500 and DX10. R300 beat out DX9 by a few months, but DX10 is by general assumption supposed to coincide with Longhorn, and Longhorn itself keeps getting bumped back... If R500 parts are actually going to be out in late-2004/early-2005 but Longhorn itself isn't going to be out until 2006...? Are ATi and nVidia going to be beating out the DX10 spec by THAT much? X2's launch likely wouldn't be that far off, but if the PC parts using those architectures are out much earlier it seems like a LONG time to wait for DX10 to start cropping up. And especially, would they be wanting to delay X2 developers from using DX10 until it's fully realized on the PC?

Granted there's a lot of assumptions intermingling, but the timing seems to be off. I'm thinking at this point we may well see DX10 (or a close equivalent still in the DX9 domain) before Longhorn, because I don't think they want to stunt its development by tying it to their major OS overhaul that is already lagging and may have to even more before they're satisfied with its launch state.

 

[Wunderchu]

Seems fine to me, though I'm really starting to wonder about the delay between R500 and DX10. R300 beat out DX9 by a few months, but DX10 is by general assumption supposed to coincide with Longhorn, and Longhorn itself keeps getting bumped back... If R500 parts are actually going to be out in late-2004/early-2005 but Longhorn itself isn't going to be out until 2006...? Are ATi and nVidia going to be beating out the DX10 spec by THAT much? X2's launch likely wouldn't be that far off, but if the PC parts using those architectures are out much earlier it seems like a LONG time to wait for DX10 to start cropping up. And especially, would they be wanting to delay X2 developers from using DX10 until it's fully realized on the PC?

Granted there's a lot of assumptions intermingling, but the timing seems to be off. I'm thinking at this point we may well see DX10 (or a close equivalent still in the DX9 domain) before Longhorn, because I don't think they want to stunt its development by tying it to their major OS overhaul that is already lagging and may have to even more before they're satisfied with its launch state.

I believe ATI is now on an 18 month product cycle (major release every 18 months, with refresh parts released at 6 month intervals inbetween)

.. if this 18 month product cycle holds true going forward, then ATI 's release schedule should be something like the following:

R420 -- released early spring, 2004
(R420 refresh) -- released early fall, 2004
(R420 refresh #2) -- released early spring, 2005
R500 -- released early fall, 2005


... if something like that release schedule does holds true going forward, then R500 should be out in late 2005, which would not be too early, compared to Windows Longhorn's and DirectX 10 's release....

 

[Vince]

Does anyone know what this 4.0 version entails over 3.0 other than the much touted nomenclature? Which is all that people seem to care about anyways.

Personally, the addition of Topology Ops to dedicated constructs in DX is much cooler.

 

[Wunderchu]

Does anyone know what this 4.0 version entails over 3.0 other than the much touted nomenclature? Which is all that people seem to care about anyways.

Personally, the addition of Topology Ops to dedicated constructs in DX is much cooler.

Many people speculate that DirectX 10 will feature the (partial or full) integration of the 'Pixel Shaders' and 'Vertex Shaders' into (partially or fully) unified 'Shaders'

 

[cthellis42]

Dude, nomenclature is DA BOMB! I mean, just look what happens when you call Windows 95 Windows 98!

I'm just curious about the timing and the platforms; the significance will be something else entirely.

 

[nobie]

some major points in DX10:
- lots of optimization, less bus usage
- VS and PS are unified
- shaders can create and destroy geometry
- shaders can report back to the CPU (so physics can be run on the GPU, for example)

 

[London Geezer]

some major points in DX10:
- lots of optimization, less bus usage
- VS and PS are unified
- shaders can create and destroy geometry
- shaders can report back to the CPU (so physics can be run on the GPU, for example)

right for the most part.... but "lots of optimization.." ???

i always thought that's up to the programmer.....

(just teasing )

 

[Dave Baumann]

This is a firaly good indication that nearly all of the geometry ops on XBox2 are going to be pushed to the graphics processor, leaving the CPU for just AI and physics.

 

[Simon F]

Does anyone know what this 4.0 version entails over 3.0 other than the much touted nomenclature?.

I thought Megadrive's posting was a reasonable summary. You don't want to get bogged down in the details anyway.

 

[London Geezer]

-Non-real-time usage can be slow
+Performance may drop drastically below 640x480 10 Hz

what exactly does that mean???? :?

 

[Simon F]

-Non-real-time usage can be slow
+Performance may drop drastically below 640x480 10 Hz

what exactly does that mean???? :?

"Resources are unlimited"

What that means is that you can write an arbitrary length program etc and it'll run. Just don't expect the performance to be "real time" (or even proportional to the length of the code). There's probably going to be a "sweet spot" for the shader lengths etc, and if you go over it the performance will degrade rapidly.

 

[London Geezer]

-Non-real-time usage can be slow
+Performance may drop drastically below 640x480 10 Hz

what exactly does that mean???? :?

"Resources are unlimited"

What that means is that you can write an arbitrary length program etc and it'll run. Just don't expect the performance to be "real time" (or even proportional to the length of the code). There's probably going to be a "sweet spot" for the shader lengths etc, and if you go over it the performance will degrade rapidly.

mmmmmmmmm ok....... anyone else wants to try getting this into my head? cause i still don't get it..... :?

 

[DeanoC]

mmmmmmmmm ok....... anyone else wants to try getting this into my head? cause i still don't get it..... :?

What SimonF said, it will have resource virtualisation, if your shaders fit into some form of cache/on chip (whatever actual hardware the thing has) then it will run fast, but if you want to use longer shaders etc it will still run but performance may suck badly.

An example would be RAM on your PC, you can use more RAM than you machine has (upto harddrive siez), but it can run really slowly if you use that facility to much. Try using 2 Gig of RAM on a 128Meg machine, it will work but you probably can go make a cup of tea for even simple things to finish.

 

[London Geezer]

right, i get that, the thing is, how is that related to:

+Performance may drop drastically below 640x480 10 Hz

:? :? ??

thats what i dont really get
Thanx BTW!!

 

[Simon F]

right, i get that, the thing is, how is that related to:

+Performance may drop drastically below 640x480 10 Hz

My guess is that it's just a figure plucked out of the air. It's not as fast as the expected "real-time shader" behaviour, yet faster than you'd expect from a CPU-only solution. <shrug>

 

[Wunderchu]

london-boy,


What Simon F and DeanoC are trying to say is something like following:


- on current generation video cards (i.e.: DirectX 9 video cards), if you attempt to run a shader which is too long and/or too complex, it may not run at all

- however, if you try to run a super long and/or super complex shader on DirectX 10 hardware, it will run, but it may run very slowly, i.e.: 10 fps at 640x480 , or slower .........


[Note: it may become more clear if you check out around page 27 of the slides which are posted here:

http://www.beyond3d.com/forum/viewtopic.php?t=8959&postdays=0&postorder=asc&start=0

http://ilfirin.barbos.net/FF.ppt ]