I finally have the ATi documents...

[Dave Baumann]

...A few days after everyone else!

So, what do we want? Do you want a B3D tech preview like the other sites have already puiblished and attempt to clear up some of the confusing areas (I've already spotted one - colour sample compression only appears to occur in SuperSample FSAA and not Mutlisample; conversly it appears that Multisample can make use of lossless z compression but supersampling can't). Or do you just want us to wait until, by means fair or foul, we get some R300 hardware in.

If I were to do a tech preview then I could try and get some questions answered but I can't guarantee that they would be answered because dealing with ATi PR appears as effective as beating your head against a wall (sorry, ATI, but it really is like getting blood out of a stone when compared to the job NVIDIA and their agencies do!) - could I call on some of the ATi regs (OpenGL guy and ATI-ISV) to pull some strings there?

Anyway, let me know what you want guys.

[edit] The Poll will run for 1 day, whatever's highest at the end I'll base my decision on.

 

[pascal]

Dave

Why not Preview with live FAQ, it could be updated as the time flow.

 

[Randell]

Preview but concentrate on the areas other have missed.

 

[Above]

I was almost going to post a question on a related issue that occurred to me on looking at the THG preview.
I saw claims that looked as if they had been uncritically relayed from ATi, as if they merely represented the line ATi was giving. You would be able to find many such examples if you looked over it.
I wondered how far webmasters are made to comply with the wishes of company PR men in what they write, in order to obtain samples of the latest hardware. What it seems to boil down to is: Relay our line, and you will be able to publish as fast as anyone else on the latest hardware. Thus, we get rehashes of PR material because the journalists are only allowed a few hours to play with the technology but feel obliged to be among the first.
All this will seem very obvious to people and does not even suggest financial incentives. But I would rather Beyond 3D would do a proper independent review when they procure the hardware than attempt to explore hardware from PR sheets.
And of course, ATi present particular problems because features claimed on their publicity do not necessarily match what works in practice.

 

[Doomtrooper]

Beyond3D should be on the TOP 5 ATI for getting a card, at least the reviewers here are versed in 3D unlike some other brutal reviewers out there...BIG SITES don't mean better...off to hassle some ATI people .

 

[LittlePenny]

I find interviews much more revealing than press releases. I wouldn't mind hearing OpenGL guy ramble on about some of the stuff he has done or knows to be impressive. But, it wouldn't be too fair for him.

Plus I thought Kyle did a pretty good job with the ATi whitepapers, so I don't see much point in repeating all the PR and what Kyle did. Something is bothering me about that though. In that article he wrote

Also, as we see memory buffers getting bigger and bigger, we can surely be certain that we will see memory controllers getting more and more complex. Hmm, does eight controllers being used on a 256MB or 512MB RAM buffer sound excessive? Then again, maybe the advent of affordable DDRII will keep that from becoming a reality so quickly

If I am not forgetting things didn't you guys correct me on this before? Shouldn't the memory controllers always stay the same in number regardless of the amount (MB-wise)of ram?

 

[Reverend]

Beyond3D should be on the TOP 5 ATI for getting a card, at least the reviewers here are versed in 3D unlike some other brutal reviewers out there...BIG SITES don't mean better...off to hassle some ATI people .

Unfortunately, you want the biggest audience with the limited review samples availed by those decision makers from on high. That means the usual sites. Also, other IHVs know who Kristof works for and that he is a B3D staff. We can say, and you will all know, that the other B3D staff won't be swayed by any influence Kristof may feel like imparting on the B3D staff but there you go... Board vendors, on the other hand, are very different creatures.

WRT the topic at hand. It's late, it'll take a few extra days to make the pages... what would be the point anyway? Unless you can dig up additional info or have some questions answered, it's absolutely useless to make an article. There's no need for this poll/announcement - it's late, you got the docs anyway, you make an article that is better than the rest and announce it.

 

[Dave Baumann]

We can say, and you will all know, that the other B3D staff won't be swayed by any influence Kristof may feel like imparting on the B3D staff but there you go... Board vendors, on the other hand, are very different creatures.

Well, its worth pointing out, he doesn't exert any influence in the first place as he has no input on reviews. So there not even the opportunity for sway!

WRT the topic at hand. It's late, it'll take a few extra days to make the pages... what would be the point anyway? Unless you can dig up additional info or have some questions answered, it's absolutely useless to make an article. There's no need for this poll/announcement - it's late, you got the docs anyway, you make an article that is better than the rest and announce it.

I want to know if people will read on if I do one; I don't want to waste the effort for no reason, however judging from the responce so far it seems there is still a 'market' for it.

 

[Tahir2]

Yes I would like to read something to keep me entertained whilst I am twiddling my thumbs at work.

I dont mean that in a disrespectful way...

However if it doesnt add anything new then there is no point like Rev says.
And again.. if you take a different approach to the rest of the gang then there is a ready made audience anyway (i.e. members of this forum )

 

[fresh]

Bring it on Dave!

 

[alexsok]

Yes, we definetly want a preview!

 

[Basic]

I agree with Reverend, I don't see much use of a deeper analysis of the ATi press documents.

But if you'd said "bring on the questions you still got unanswered after reading other previews so we can do a Q&A", then I'd say "Here's mine". I think that would be the optimum for your effort vs our benefit.


I guess some might be more of DX9 questions than R300 questiones, so if somebody have a link to the most current DX9 spec then just post that instead.

They talk about a 128bit pixel pipeline (4x32f). Is that the native internal precision, or will it take a performance hit?

What formats have they got for textures and rendertargets?
4x8i, 10:10:10:2, 2x16i, 2x16f, 4x16f, 1x32f, 2x32f, 4x32f?

What does multiple rendertargets mean?
In one preview I saw a table that said that DX8.0 and DX8.1 can have 1 rendertarget, but DX9.0 can have 4. Does that mean that one pixelshader can write different values to 4 framebuffers in one pass?

What exactly are the rules for texture sampling?
16 texop from 16 textures + 64 colour op + 16 (dependant) texop from same textures + 64 colour op?
Or is it more flexible, with a possiblity to have more than two "phases", meaning dependant texops in more steps.

What decides the mip level on the second texture in a dependant texop?

 

[KnightBreed]

What does multiple rendertargets mean?
In one preview I saw a table that said that DX8.0 and DX8.1 can have 1 rendertarget, but DX9.0 can have 4. Does that mean that one pixelshader can write different values to 4 framebuffers in one pass?

I believe that refers to the ability to assign specific pixel shader programs to groups of vertices/textures/pixels. Rather than the program running through the entire scene, it'll only calculate for the target that has been assigned to it.

Source:

DirectX 9 calls for support for Multiple Render Targets (MRTs) which, as the name implies, means the ability to target specific objects in a scene for pixel shader programs. This method will be one of the most common in which complex pixel shader programs are implemented. You won't see 100+ instruction programs being used on every single object in a scene, but in order to make a few parts of the scene truly stand out MRTs will come in handy

 

[Basic]

Uhmm, he implies that the 'target' in "multiple render targets" is specific objects in the scene. That doesn't make sense to me, you do that by switching to another shader. I'm sure that it in some way or another refers to frame buffers/aux buffers/textures.

 

[LeStoffer]

Why not make it easier for yourself? Just make the documents available as a download for all of us!

 

[Pete]

If you want to throw a little article together to clear up issues you feel have been neglected or incorrectly addressed, I say go for it. You might make it a technology review, and get that out of the way until you get hardware to benchmark. Two smaller articles might be better for the readers and the server.

But if you'll still need the hardware to make definitive conclusions, you might be better off waiting.

 

[OpenGL guy]

Uhmm, he implies that the 'target' in "multiple render targets" is specific objects in the scene. That doesn't make sense to me, you do that by switching to another shader. I'm sure that it in some way or another refers to frame buffers/aux buffers/textures.

When you think about "multiple render targets" think GL_FRONT_AND_BACK rendering. I.e. one Z compare, multiple output pixels.

 

[Basic]

Thanks OpenGL guy. Especialy for the last sentence, since the first actualy could be interpreted in two ways.

The constant GL_FRONT_AND_BACK is used with two completely different meanings. One with glMaterial and its likes, where it means front and back face of a poly. And another meaning with glDrawBuffer, where it means front and back buffer.

From your last sentence I guess you mean the 'glDrawBuffer' meaning. Should it be interpreted as the same pixel to many buffers (as in OGL), or have they added the possibility to generate different values to different outputs.

I (oddball as I am) would actually like the idea to run a couple of image filters simultaneously on the same input image(s). Or to put it another way, I want to do filters where the output is a vector with more than four elements.

 

[Sharkfood]

I'd say wait until you get the card.

We all know what confusion and expectation level was created with the R200 from early tradeshow information, early PR rumors and early documents concerning the chipset. These have a tendency to change between the initial product publicity release and the time the unit hits a box and sits on a shelf.

I think the most value can be obtained by seeing actual reproduced results on the physical piece of hardware, whereas PR documentation creates more confusion, wrongful guesses on what they exactly mean by some of their terminology, missed expectations, the occasional flame war or similar - which is more damaging than good.

 

[KimB]

My feelings:
Unless you feel you have something to say that nobody yet on the other websites has said, I think it would be pointless (from where I'm sitting).

Granted, as far as the website is concerned, it might be a good idea to go ahead and do it for a few more hits.