Is Doom 3's release going to change the OpenGL community?

[Richard]

The point being that this is still work that has been done - its even worse that its now wasted work.

I'm not disputing that the ARB is ineffective when compared to Microsoft. Ruling by comittee is always a pain. I'm simply pointing out that according to Robert Duffy there are basically two "paths" in DOOM 3: ARB2 and 'everything else'. Getting rid of the ARB path is just another way of showing "you can play the game at full glory with the ARB2 path or using hacks and missing effects for the vendor-specific ones".

Btw, since he commented on the P10 once, I wouldn't mind a short Q&A organized by B3D with JC about the P20.

And which of these do you think eg. the 7x00 Radeons will run...?

The magazine states the GF4mx and Radeon 8500 are the minimum supported cards.

 

[Joe DeFuria]

So, the only non DX9 cards that will run Doom3 are from ATI and nVidia?

 

[PowerK]

So, the only non DX9 cards that will run Doom3 are from ATI and nVidia?

Hmm.. that's a good question.
Was Matrox Parhelia DX8 or DX9 level ? I'm not even sure if this Matrox cards are supported.

 

[Richard]

So, the only non DX9 cards that will run Doom3 are from ATI and nVidia?

Well, there used to be a Parhelia path but we haven't heard anything about that for a long time. Could have gone the way of the dodo like the ARB(1) path. Comparing general Parhelia's performance to, say, the Radeon 8500 might shed some light if it's still supported or not.

Hopefully, id Software will release more information soon enough.

 

[anaqer]

So, the only non DX9 cards that will run Doom3 are from ATI and nVidia?

I was under the impression that it's been like this from the beginning.
I kinda wonder how the "new" S3 cards will handle Doom3 though.
(BTW Parhelia is some sort of hybrid , IIRC "almost-VS2.0" + PS1.x )

 

[nutball]

It strikes me that with DX10 being so closely tied in to Longhorn (2006?), that the ARB might have a bit of a breathing space to develop OGL to a point where it will at least equal DX10 when that arrives.

If I've interpreted what I've read here over the past many months correctly, the next-gen cards from the two big boys (NV50, R500) will come out quite some time before DX10. Maybe even a whole generation?

So who's going to support all the wonderful new features for them? DX9.0d? DX9.1?

 

[PeterAce]

It strikes me that with DX10 being so closely tied in to Longhorn (2006?), that the ARB might have a bit of a breathing space to develop OGL to a point where it will at least equal DX10 when that arrives.

If I've interpreted what I've read here over the past many months correctly, the next-gen cards from the two big boys (NV50, R500) will come out quite some time before DX10. Maybe even a whole generation?

So who's going to support all the wonderful new features for them? DX9.0d? DX9.1?

Anything glaringly wrong with this :

Looking at the landscape untill DirectX Next id guess the Graphics scheduals are somthing like :

ATI:

R420 AGP - Spring 2004
R423 PCI-E - Fall/Christmas 2004
R480 AGP - Fall/Christmas 2004 (R420 Refresh)

R520 (SM 3.0 +) - Spring/Fall 2005

R600 (SM 4.0 and DirectX Next features) - Fall/Christmas 2006

NVIDIA :

NV40 AGP - Spring 2004.
NV45 PCI-E - Fall/Christmas 2004.
NV48 AGP - Fall/Christmas (NV40 Refresh)

NV50 (SM 3.0 +) - Fall/Christmas 2005

NV60 (SM 4.0 and DirectX Next features) - Fall/Christmas 2006

The DirectX Next cards will prob be a "Unified Vertex/Pixel Shader design" basicly a pool of re-allocatable ALU units.

The "+" sign denotes FP Frambuffer/Blend with MSAA working with them.

 

[eSa]

I not claiming to know anything but I find really hard to believe that ATI would wait with "revolutionary R500/R520" to the fall 2005. I think "shader model 3.0" part will be coming out in spring 2005. Maybe even in early spring.

If Nvidia will get the clocks up for the 6800 series refresh, they will have really viable contender for X800 also in performance wise... But you all have heard that million times.

 

[GameCat]

Ingenu, the problem is that the ARB can just be too slow to react in making changes to the core - while the ARB fragment and vertex shader extensions are now available, how long did it take the ARB to settle on them? IIRC the vertex program extension was available for some time before the arb finally hammered out a version of the fragment program extension that they could all settle on, and this was long after DX9 was released.

Actually, ARB_fragment_program was available before any hardware supported it and long before DX9.

The only thing the ARB hasn't standardised in OpenGL are pixel shader assembler extensions for DX8 level hardware and assembler vertex and pixel shaderes for DX9 level hw. The reason assembler extensions hasn't been created for DX9 level hw is obviously that the ARB wanted to focus on glsl instead.

There really isn't that big of a need to standardise a vast number of extensions, mostly it's already been done. 2.0 should fix a number of outstanding isues like handling float data types,, but I can't really think of any area where modern hardware offers wildly differing interfaces to the same functionality.

 

[PowerK]

ATI:

R420 AGP - Spring 2004
R423 PCI-E - Fall/Christmas 2004
R480 AGP - Fall/Christmas 2004 (R420 Refresh)

R520 (SM 3.0 +) - Spring/Fall 2005



NVIDIA :

NV40 AGP - Spring 2004.
NV45 PCI-E - Fall/Christmas 2004.
NV48 AGP - Fall/Christmas (NV40 Refresh)

NV50 (SM 3.0 +) - Fall/Christmas 2005

NV60 (SM 4.0 and DirectX Next features) - Fall/Christmas 2006

An interesting thing is that currently, ATI X800XT PE (and/or NVIDIA 6800Ultra) is almost non-existent in the market.

It very much looks like they will *start* becoming readily available from mid august ~ early september. Do you think ATI (or NVIDIA) is giving these current flagship cards only 2~3 months *maximum* of their life-time until they release refreshs (R480 and/or NV48) ?

I think this whole delay fiasco (be it ram shortages, poor yields on their VPU or whatever...etc) has a very good chance of affecting their future products in their roadmap. They might end up delaying releases for refreshs (R480 and/or NV48) hence another delay to the next generation (R520/NV50) releases, or they might go straight to R520/NV50.

I think this delay with X800XT PE and/or NVIDIA 6800Ultra means something changed in their near future plan as well.

 

[keegdsb]

So, the only non DX9 cards that will run Doom3 are from ATI and nVidia?

Well, there used to be a Parhelia path but we haven't heard anything about that for a long time. Could have gone the way of the dodo like the ARB(1) path. Comparing general Parhelia's performance to, say, the Radeon 8500 might shed some light if it's still supported or not.

Hopefully, id Software will release more information soon enough.

No, there never was a Parhelia path. Or, if he started one, it was never completed. I have emailed Carmack about Parhelia and his response did not inspire confidence.

It can "in theory" run the (now defunct?) ARB path, "but performance was really bad." He also said there were problems due to hardware bugs.

Mr. Carmack hasn't even talked to Matrox in "2 years or so."

Is the ARB path going to be completely removed or just unsupported? I would at least like a shot at running it on my system. :?

 

[Cryect]

Heh, well Carmack does seem to have a tendency of getting pissed off at companies for having OpenGL bugs. As when he talked about in his plan not talking to ATI for several months after a beta Radeon 8500 wouldn't render the console correctly

I could imagine if there were severe hardware bugs he wouldn't even bother considering looking at a piece of hardware. Can't blame him since why waste time on something that doesn't work correct to spec and is such a minor portion of potential consumers that its not worth bothering over due to this.

 

[aZZa]

Was there ever a P9/P10 (3dLabs) path or Kyro (1/2) path ever developed or in development????

Would a PVR Series 5 chip - earn a new path or default ARB2 be sufficient??? -if released - not enough development money on offer to bother?? No personal attack on Id for this reply - but would we see a patch ala Far Cry - if a card came out.

Volari & Deltachrome gotta run it either via arb2 or minimum at r200 path for ok quality.

 

[Richard]

No, there never was a Parhelia path. Or, if he started one, it was never completed. I have emailed Carmack about Parhelia and his response did not inspire confidence.

That which cannot be named has a Parhelia path. Whether it was even functional I don't know however, JC's June 25 2002 .plan update says the card will run DOOM. Like I said, a lot could have changed since then. Can you tell us when you talked to JC?

Is the ARB path going to be completely removed or just unsupported? I would at least like a shot at running it on my system. :?

Here's what Robert Duffy recently said about the paths:

NV10 - GF4MX etc.
NV20 - GF3
ARB - this path is not used.
R200 - ATI 8500 and a few newer models that are essentially faster R200's
ARB2 - used for modern NVIDIA and ATI parts.

 

[anaqer]

R200 - ATI 8500 and a few newer models that are essentially faster R200's

Them being...?

 

[Xmas]

R200 - ATI 8500 and a few newer models that are essentially faster R200's

Them being...?

Radeon 9000, 9100, 9200

 

[Xmas]

Was there ever a P9/P10 (3dLabs) path or Kyro (1/2) path ever developed or in development????

P9/P10 can run the NV20 (or NV10?) path, AFAIK. And Kyro has no specific extensions.

Would a PVR Series 5 chip - earn a new path or default ARB2 be sufficient???

If S5 makes it to the PC, it can use the arb2 path, like all other DX9 level cards.

Volari & Deltachrome gotta run it either via arb2 or minimum at r200 path for ok quality.

They can't run the r200 path, but they should be able to run the arb2 path.

 

[Rolf N]

Actually, ARB_fragment_program was available before any hardware supported it and long before DX9.
<...>
The only thing the ARB hasn't standardised in OpenGL are pixel shader assembler extensions for DX8 level hardware and assembler vertex and pixel shaderes for DX9 level hw.

Huh?
ARB_fragment_program is a "DX9 level" "pixel" shader assembly extension.
What's missing is an ARB extension that exposes control flow in vertex shaders.

P9/P10 can run the NV20 (or NV10?) path.

Vanilla NV_register_combiners code runs fine on the P9, but it doesn't support NV_texture_shader, and it doesn't support EXT_fogcoord. If the NV20 path requires one of these, it won't work out (but I don't know).

If the NV10 path requires EXT_fogcoord, it won't even be able to run that.

Volari & Deltachrome gotta run it either via arb2 or minimum at r200 path for ok quality.

They can't run the r200 path, but they should be able to run the arb2 path.

Current DeltaChrome drivers don't support ARB_fragment_program. I agree that they should
I haven't heard anything at all yet about the quality of XGI's ARB_fp implementation :?

edited bits: two and still counting

 

[Humus]

IIRC the vertex program extension was available for some time before the arb finally hammered out a version of the fragment program extension that they could all settle on, and this was long after DX9 was released.

This myth continues to live on I see. There's no more clear point where OpenGL was just way ahead of DirectX than around the DX9 release. I released 5 demos using GL_ARB_fragment_program before DX9 was released.

GL_ARB_fragment_program was approved by ARB on September 18, 2002.
DirectX 9 was released on December 19, 2002.

The only reason people think DirectX is ahead is because it's hyped up at every new version.

 

[karlotta]

Will ATIs new OGL driver change the industry.....will it be out before Doom3, or right after with the august driver release?