[g__day]

We have all seen the NV3X and R3XX cards and know they were called fully Directx 9.0 compliant if I recall correctly. They were not labelled Directx 9.0a or b compliant but DX9.0 compliant infering the superset of all DX9.0 anything releases!

Now DX9 contains shader model 3.0 and NV3X and R3XX aren't fully shader model 3.0 compliant by a long strech.

So is it incorrect or misleading advertising to claim these cards as fully DX9 compliant?

* * *

http://www.ati.com/products/radeon97...pro/index.html
Overview
RADEONâ„¢ 9700 PRO is the world's fastest* and most advanced graphics board, featuring ATI's RADEONâ„¢ 9700 PRO Visual Processing Unit (VPU). Through a combination of incredible 3D rendering performance, sophisticated real-time visual effects, unsurpassed image quality and cutting-edge video features, it takes the PC entertainment experience to a totally new level.

At-a-glance:
Fastest* 3D gaming performance with next-generation VPU architecture
Complete DirectX® 9.0 support for unprecedented realism and sophisticated visual effects

* * * Its that word 'complete' support that makes me wonder...

I had infered:

DX9 compliant -> has some DX9 only functionality from the 100s of DX9 only attributes

DX9 compliant in hardware -> has some DX9 functionality that is fully done in h/w

DX9 fully compliant in h/w -> has all DX9 functionality fully revealed and fully supported in hardware.

Now the confusing thing is R3XX is considered a full DX9 card - yet DX9 reveals not only shader model 2.0 but soon shader model 3.0. R3XX certainly isn't shader model 3.0 compliant so why is it considered a "fully" compliant DX9 card?

Not sure about the namining conventions - can anyone explain it authoritively please!

[Wixard]

dx9.0b for example, is not the same version of dx as dx9.0

I guess you could argue that saying its dx9 compatabile soft-of infers dx9.0xxx. I dunno, email them and tell them

[MoboJack]

R3xx and Nv3xx cards are DirectX 9.0b complaint. Microsoft is to launch 9.0c with Ps3.0 and a lot more features.

[fallguy]

Agreed. As long as they say DX 9.0, I dont see how they could be wrong. The average consumer wont know the difference though, so it could be considered shady by some.

But then on the other side of the fence, when running in FP16.. thats not DX9 either.

[Evildeus]

Originally Posted by fallguy

But then on the other side of the fence, when running in FP16.. thats not DX9 either.

Since when?

[I.S.T.]

It's not full DX9 if FP16 is ran all the time, rather than just part of the time, IIRC.

[Evildeus]

Sure but in partial precision, FP16 is DX9.0.

[g__day]

My concern is DX9 compliant or DX9.0 compliant should cover all variants that fall under that name. Not we cover a and b but not c - that is the mis-leading part.

Perhpas they thought a DX9.1 would come out and not leave them exposed for false advertising. But given its still a DX9.0 variant that delivers full shader model 3.0 exposure - I say they have a headache IMHO.

[Quitch]

Will PS 3.0 be a required part of the spec in c?

[gokickrocks]

Originally Posted by Quitch

Will PS 3.0 be a required part of the spec in c?

it will be part of it, but it wont be required

[bloodbob]

Well this is where everything goes to hell and some of the community backs it up.

They say that DX9.0c is just a new runtime/compile time libaries yet at the same time this run time no longer conforms to the orginal standard.

Now PS3.0 has always been part of the standard just never exposed in the run time. But with the libaries that now exposes it they also decided to change the standard as such.

I guess their is DX9.0 complaint DX9.0 compatible and DX9.0 something ( I can't remeber which ones goes here ) with DX9.0C SM 3.0 support.

Complaint and compatible mean to different things one of them says it meets some driver interfece standard even though the hardware can be like DX7.0 and another one requires the hardware support.

[Sandwich]

Originally Posted by bloodbob

Complaint and compatible mean to different things one of them says it meets some driver interfece standard even though the hardware can be like DX7.0 and another one requires the hardware support.

As I understood it. Compatible means the cards got drivers that work with that version of directx.
Complaint means it can actually do the bells and whistles that are minimum requirement set by MS to still be 'complaint'.

So dx9 compatible hardly means anything.

[Hellbinder]

fp16 is part of DX9. A pure FP16 card would still be considered DX9 complient (at least i think so)

Which is why i suggest that the r420 could be a lot closer to SM 3.0 then some poeple are suggesting. FP24 now becomes what FP16 was. If instruction processing is increased and some other capabilities added it may not *quite* make PS3.0 to a tee but would be virtually the same thing as for all practical purposes.

Just a thought.

[Evildeus]

Originally Posted by Hellbinder

fp16 is part of DX9. A pure FP16 card would still be considered DX9 complient (at least i think so)

No, you need to be at least FP24. _pp is an option only.

[Chalnoth]

Originally Posted by Hellbinder

fp16 is part of DX9. A pure FP16 card would still be considered DX9 complient (at least i think so)

Well, even if it were possible, such an architecture would have texturing errors left and right.

[BRiT]

Originally Posted by Chalnoth

Well, even if it were possible, such an architecture would have texturing errors left and right.

*cough*NVidia 5200*cough* Or is it more FX than FP?

[Neeyik]

5200 = FP32 tex addr-cum-ALU plus FX12 ALUs.

[hoom]

Um if you'd all just re-read the "M$ clarifies directx9.0c vs 9.1" stuff, sm3.0 is actually already exposed in the 9.0b runtime, while sm3.0 support for developers is in 9.0c sdk.
9.0c (& 9.0d if needed) runtime is being reserved for when hardware comes out (nowish) & if there is an incompatibility between the previously written sm3.0 functionality in 9.0b runtime and the real hardware.

Also, stop thinking of the runtime and the SDK as one thing, ie. the runtime shipped in "DirectX SDK Update (Summer 2003)" is 9.0b. This was the latest released runtime (i.e. you could get this runtime, including the developer runtime, publicly whilst the SDK was still in beta). This enables us to release the SDK on a significantly more frequent basis than runtime changes...

Some of the confusion may be due to us shipping support for 3.0 shader models in the DirectX 9 runtime. This was specifically to reduce the amount of churn in the API. When some 3.0 hardware is produced we'll evaluate how close the API fits the hardware (there's always some tuning), and we may need to release an update - but remember the definitions, this would most likely be a "letter" change runtime release, not a "minor".

http://www.beyond3d.com/forum/viewto...;view=previous

[added quote/link] which sounds like there was sm3.0 support in 9.0 runtime even though the sdk didn't until the 9.0c sdk.

[rainz]

Originally Posted by g__day

We have all seen the NV3X and R3XX cards and know they were called fully Directx 9.0 compliant if I recall correctly. They were not labelled Directx 9.0a or b compliant but DX9.0 compliant infering the superset of all DX9.0 anything releases!

Now DX9 contains shader model 3.0 and NV3X and R3XX aren't fully shader model 3.0 compliant by a long strech.

So is it incorrect or misleading advertising to claim these cards as fully DX9 compliant?

* * *

http://www.ati.com/products/radeon97...pro/index.html
Overview
RADEONâ„¢ 9700 PRO is the world's fastest* and most advanced graphics board, featuring ATI's RADEONâ„¢ 9700 PRO Visual Processing Unit (VPU). Through a combination of incredible 3D rendering performance, sophisticated real-time visual effects, unsurpassed image quality and cutting-edge video features, it takes the PC entertainment experience to a totally new level.

At-a-glance:
Fastest* 3D gaming performance with next-generation VPU architecture
Complete DirectX® 9.0 support for unprecedented realism and sophisticated visual effects

* * * Its that word 'complete' support that makes me wonder...

I had infered:

DX9 compliant -> has some DX9 only functionality from the 100s of DX9 only attributes

DX9 compliant in hardware -> has some DX9 functionality that is fully done in h/w

DX9 fully compliant in h/w -> has all DX9 functionality fully revealed and fully supported in hardware.

Now the confusing thing is R3XX is considered a full DX9 card - yet DX9 reveals not only shader model 2.0 but soon shader model 3.0. R3XX certainly isn't shader model 3.0 compliant so why is it considered a "fully" compliant DX9 card?

Not sure about the namining conventions - can anyone explain it authoritively please!

Are you trying to get somes serious arguments for a free card ? hehe

RainZ

[digitalwanderer]

Oh man, this is gonna get ugly before it's over! (I'm refering to the whole dual launch thing, not this thread. )

[g__day]

No, I am saying many countries labelling laws impose still penalties for mis-leading advertising. This means consumer watch-dogs are on alert for deceptive practices - both by intent or mis-adventure.

With the inclusion and soon enablement of shader model 3.0 within Directx 9.0 it raises the issue of "What is Directx 9.0 compliance?" If as a distributor or manufacturer I claim Directx 9.0 compliance - but its actually Directx 9.0 compatability - I have only partially delivered on my promise when I sell shader model 2.0 cards. So the key question is - have I mis-represented my products?

Marketing needs clearer, more precise language to define what is being delivered. Perhaps to say DX9 shader model 2.0 or DX 9 shader model 3.0 compliance. Then their Web factsheets, marketing material and cards packaging material needs to be updated to ensure this more precise message is delivered.

To the extent that this is not done, this might inflate the risk of a watch dog biting the major card providers.