WinHEC slides give info on early spec of DX in longhorn

[psurge]

Sage, Slide 18:

Texture filtering and alpha blending are allowed to take some shortcuts.

 

[Sage]

yes, I read the entire thing.

 

[radar1200gs]

I'd like to draw peoples attention to slide number 11 in TW04079_WINHEC2004.ppt, and in particular to the line that reads:

"IEEE compliance for floating point"

I'd like to invite those who in the past have argued in these forums that IEEE FP compliance is unnecessary and a waste of performance/transistors to justify their arguments in the light of the above slide. I'd also like them to address their statements that certain cards would continue to be more capable than others in the future.

Also given that the average consumer purchases a machine and uses it for 3 years before disposing of it or upgrading it significantly would you say that a GF6800 based card or a Radeon series card would be a better choice for such a consumer?

I would like to draw your attention to the line on slide 12 that reads:

"IEEE 754 compliance (well, sort of)"

I believe IEEE 754 covers FP16, not FP24

 

[Dave Baumann]

Also given that the average consumer purchases a machine and uses it for 3 years before disposing of it or upgrading it significantly would you say that a GF6800 based card or a Radeon series card would be a better choice for such a consumer?

It makes little difference. The slides are talking about the next generation API (i.e. Direct Next) of which neither 6800 or X800 are going to be compliant with. These particular slides are not talking about the requirements for Longhorn, just DX Next, which will still support legacy DX hardware, such as 6800 and X800, through compatibilty with previous API models. Slide 6 covers the changes to DirectX9 specific for Longhorn, and any hardware that can cope with these will be OK with the OS.

 

[radar1200gs]

Also given that the average consumer purchases a machine and uses it for 3 years before disposing of it or upgrading it significantly would you say that a GF6800 based card or a Radeon series card would be a better choice for such a consumer?

It makes little difference. The slides are talking about the next generation API (i.e. Direct Next) of which neither 6800 or X800 are going to be compliant with. These particular slides are not talking about the requirements for Longhorn, just DX Next, which will still support legacy DX hardware, such as 6800 and X800, through compatibilty with previous API models. Slide 6 covers the changes to DirectX9 specific for Longhorn, and any hardware that can cope with these will be OK with the OS.

While neither may be fully compliant, one will come closer than the other. If a geometry shader is all that is needed to complete the picture, then NV40's vertex shaders can already act as a tesselator in hardware by using two different speed loops. R420 can't do this because the CPU has to do its vertex looping for it.

 

[jvd]

While neither may be fully compliant, one will come closer than the other. If a geometry shader is all that is needed to complete the picture, then NV40's vertex shaders can already act as a tesselator in hardware by using two different speed loops. R420 can't do this because the CPU has to do its vertex looping for it.

your either compliant or your not .

Doesn't matter how close to it you are. If your not compliant then it doesn't matter.

Face it. Its a new api and neither of these cards will be closer than another to it .

 

[jimmyjames123]

The slides are talking about the next generation API (i.e. Direct Next) of which neither 6800 or X800 are going to be compliant with.

What type of hardware requirements will it take to be compliant with "DirectX Next"?

 

[jimmyjames123]

your either compliant or your not .

Doesn't matter how close to it you are. If your not compliant then it doesn't matter.

Face it. Its a new api and neither of these cards will be closer than another to it .

You are just arguing semantics here. There are always degrees of compliance, and partial compliance is better than little to no compliance at all.

 

[Dave Baumann]

While neither may be fully compliant, one will come closer than the other. If a geometry shader is all that is needed to complete the picture, then NV40's vertex shaders can already act as a tesselator in hardware by using two different speed loops. R420 can't do this because the CPU has to do its vertex looping for it.

They are either compliant or not compliant, and you can be rest assured that neither will be - they don't have full virtualistion, which is whats needed. The diagrams also indicate that a geometry shader and tesselator are required - they are two separate things; also NV40's VS has no capabilities for creating geometry, as illustatred by the fact that detailed meshes have to be given in order to support Displacement mapping.

 

[Dave Baumann]

You are just arguing semantics here. There are always degrees of compliance, and partial compliance is better than little to no compliance at all.

No, this is just adhearing to the spec - technically anything is "partially compliant" with future DX capabilites. Just like R420 is "partially compliant" with SM3.0, it doesn't mean that it can utilise SM3.0 compiled shaders.

As for the requirements, read the article linked to.

 

[jvd]

your either compliant or your not .

Doesn't matter how close to it you are. If your not compliant then it doesn't matter.

Face it. Its a new api and neither of these cards will be closer than another to it .

You are just arguing semantics here. There are always degrees of compliance, and partial compliance is better than little to no compliance at all.

Great. Would u like to eat at a resturant that is only compliant with some things the board of health wants them to be at ?

You think they'd pass the health examinations.

 

[FUDie]

I believe IEEE 754 covers FP16, not FP24

Nope, it doesn't cover either.

-FUDie

 

[BRiT]

Maybe because it would require them to actually do something? Their moto seems to be sloth. They haven't done anything with their technology in at least the past 5 years.

Actually the reason I was talking to them was related to just that... good things ahead.

It's about damn time we saw some sort of advancement in sound from Creative. Everyone should be tired of Creative passing off reverb/echo as some major technical advancement under the guise of catchy marketting terms like EAX-4. They produce sound decelerators instead of accelerators. I'd be happy if they merely remarketed A3D.

 

[radar1200gs]

I believe IEEE 754 covers FP16, not FP24

Nope, it doesn't cover either.

-FUDie

IEEE 754 : 7+5+4 = 16

 

[radar1200gs]

your either compliant or your not .

Doesn't matter how close to it you are. If your not compliant then it doesn't matter.

Face it. Its a new api and neither of these cards will be closer than another to it .

You are just arguing semantics here. There are always degrees of compliance, and partial compliance is better than little to no compliance at all.

Great. Would u like to eat at a resturant that is only compliant with some things the board of health wants them to be at ?

You think they'd pass the health examinations.

Look at it this way, the nVidia restaurant will serve a dish that is 90% compliant with health standards, the ATi restaurant will seve a dish 50% compliant with health standards.

 

[jimmyjames123]

No, this is just adhearing to the spec - technically anything is "partially compliant" with future DX capabilites.

Obviously the current generation (ie NV4x and R4xx) architectures will not be "fully" compliant with DirectX Next, but that doesn't mean that one architecture does not represent a better vision of DirectX Next than the other. One can only assume that the NV4x is closer to that vision. Again, this is just a semantical argument.

Just like R420 is "partially compliant" with SM3.0, it doesn't mean that it can utilise SM3.0 compiled shaders.

Like I said, there are degrees of compliance. I don't think that most people would consider the R420 to have a very high level of compliance at all with SM 3.0, considering that it does not have full support for VS 3.0 or PS 3.0.

A few months ago, there was speculation that the R420, while not having full support for SM 3.0, would have full support for VS 3.0. Most people believed that VS 3.0 was the most important element of SM 3.0, so they claimed that this partial compliance to SM 3.0 was plenty good. Now, people are arguing that it is either "full" compliance or "no" compliance, period. How interesting.

 

[Sage]

your either compliant or your not .

Doesn't matter how close to it you are. If your not compliant then it doesn't matter.

Face it. Its a new api and neither of these cards will be closer than another to it .

You are just arguing semantics here. There are always degrees of compliance, and partial compliance is better than little to no compliance at all.

Great. Would u like to eat at a resturant that is only compliant with some things the board of health wants them to be at ?

You think they'd pass the health examinations.

Look at it this way, the nVidia restaurant will serve a dish that is 90% compliant with health standards, the ATi restaurant will seve a dish 50% compliant with health standards.

first of all, those numbers are not accurate. secondly, i wouldn't be eating either one of them! would you rather eat a cockroach or a cockroach that just took a turdbath?

 

[jvd]

Look at it this way, the nVidia restaurant will serve a dish that is 90% compliant with health standards, the ATi restaurant will seve a dish 50% compliant with health standards

and both would be closed down and wouldn't be allowed to serve anything because they both failed.

I would only eat with a dish that is a 100%.

 

[Dave Baumann]

Obviously the current generation (ie NV4x and R4xx) architectures will not be "fully" compliant with DirectX Next, but that doesn't mean that one architecture does not represent a better vision of DirectX Next than the other.

But when talking about specification of futre requirements having a "better vision" makes little difference if that vision can't achieve compliance.

A few months ago, there was speculation that the R420, while not having full support for SM 3.0, would have full support for VS 3.0. Most people believed that VS 3.0 was the most important element of SM 3.0, so they claimed that this partial compliance to SM 3.0 was plenty good. Now, people are arguing that it is either "full" compliance or "no" compliance, period. How interesting.

Not really, not in relation to this topic.

 

[jimmyjames123]

But when talking about specification of futre requirements having a "better vision" makes little difference if that vision can't achieve compliance.

Let me put it more clearly: the NV40 has a featureset that is more in tune with DirectX Next than the R420, period. That is a fact. A fact can be downplayed, but it is a fact nonetheless. I think it has already been clearly established that both the NV40 and R420 will not be "fully" compliant with DirectX Next.