3Dc
- Thread starter vnet
- Start date
Even if that one only works on a small percentage of video cards, and therefore doesn't benefit the vast majority of gamers? You are basically saying that marketing divisions should decide on features, not standards committees, and this is A Bad Thing.jvd said:
Even if you (selfishly) say, "I'm going to buy an X800, so screw the people who don't get one", you are still being short-termist. What happens if Nvidia trounce ATI in the next round (yeah, I know it's unlikely, but then so was what happened last round)? Will 3DC still be being pushed then? Surely it's always better to use a standard, if one exists?
The whole point of API's is that they should be hardware neutral - this benefits everyone the most. Whether it's Nvidia, ATI or whoever, proprietry technology should be presented to working groups for aproval, not just implemented unilaterally. Remember Glide? Let's not go back there...
Dio said:
I have no objections towards that or Ati. Especially because of the whitepaper. (I somehow doubt that NVidia would have done something like that if they came up with 3Dc, although i don't know when the whitepaper was produced). Rather the opposite. You release a new feature, it's free for all to use and you say that DXT5 is a good fallback (that's how i interpret the whitepaper at least) for now.
Most of the objections was towards people that seems to have changed their minds rather drastically as of late when it comes to usage of "features released by one IHV that doesn't come from MS"
Especially since the said feature supposedly could easily be supported on other cards, although with somewhat lower quality. (remains to be seen how big the difference is though).
Diplo:
Glide and 3Dc is not a valid comparision. Ati is making 3Dc an open standard while I if I remember correctly 3dfx sued Nvidia back in mid-1998 after they added a glide wrapper to their drivers. No one is being held back if anyone and everyone is free to implement it in later cards, and that's the important difference from Glide.
You could make a just as valid argument that NV consumers are being selfish by buying a 6800 and are in essence saying "screw you" to those cards that don't support V/PS3.0.
EDIT: Grammatical error fixed.
Glide and 3Dc is not a valid comparision. Ati is making 3Dc an open standard while I if I remember correctly 3dfx sued Nvidia back in mid-1998 after they added a glide wrapper to their drivers. No one is being held back if anyone and everyone is free to implement it in later cards, and that's the important difference from Glide.
You could make a just as valid argument that NV consumers are being selfish by buying a 6800 and are in essence saying "screw you" to those cards that don't support V/PS3.0.
EDIT: Grammatical error fixed.
Joe DeFuria
Legend
akira888 said:
Almost correctly...they sued Creative Labs.
jvd said:
Exactly. Like I said in my previous posts, even if 3Dc is only a marginal improvement over DXT5, it will be a better if not only by the virtue of it being promoted. Sure, the 'fair' way to play it would be to use DXT5 since it has a larger support base - but in ATI's view, it's competition. They will push the leverage they have to gain an advantage over the competition. And if this means pushing 3Dc over DXT5, even if the benefit is only marginal, so be it. Nothing wrong with that.
I use the same argument to support and welcome nVidia's promotion of PS3.0. Many people here criticize them for pushing PS3.0 and saying, "Uh, no! It's only a marginal improvement! It's just nVidia trying to make their product look better!" Well, duh... that's the point. They're going to leverage what advantages they have, just like ATI. Again, nothing wrong with that. That only covers the 'moral' aspect of it all, however (which folks keep tripping all over, it's irrelevant IMO). From a pure technical stand point, I would like to see a comparison between 3Dc and DXT5 just to satisfy my, and everyone elses, curiosity.
Bjorn said:Most of the objections was towards people that seems to have changed their minds rather drastically as of late when it comes to usage of "features released by one IHV that doesn't come from MS"
Kinda looked the other way around to me. Those that are always definding this type of action as "just business" appeared to be the ones started off against 3Dc and requiring fallbacks. Guess its just the way you look at it really.
Joe:
Yes you're certainly correct, I remember now - it was Creative.
Bjorn:
I strongly disagree with not adding a DXT5 normal map compression fallback for non-R4xx owners. But 3Dc does seem (from Ati's pics at least) to be somewhat more representionally accurate to the non-compressed normal map and I'll defend it on that merit. But if DXT5 compression is adequate for normal/height maps then why do so few developers seem to have used it?
Diplo:
Since it's an open standard I would not call it "proprietary" as glide was, at worst it's "temporarily exclusive." But this isn't the first time - F-buffer, NV40's video unit, truform, and the NV10 register combiner all exposed functionality not found in the current APIs or beyond what the API provided.
Yes you're certainly correct, I remember now - it was Creative.
Bjorn:
I strongly disagree with not adding a DXT5 normal map compression fallback for non-R4xx owners. But 3Dc does seem (from Ati's pics at least) to be somewhat more representionally accurate to the non-compressed normal map and I'll defend it on that merit. But if DXT5 compression is adequate for normal/height maps then why do so few developers seem to have used it?
Diplo:
Since it's an open standard I would not call it "proprietary" as glide was, at worst it's "temporarily exclusive." But this isn't the first time - F-buffer, NV40's video unit, truform, and the NV10 register combiner all exposed functionality not found in the current APIs or beyond what the API provided.
Dio said:
I am 100% behind this view Dio, and I commend ATI for calling attention to the normal map issue, doing some research, coming up with a simple low-cost solution for older cards (the alpha channel trick), plus taking it once step further by innovating a method to provide equal quality to both tangent component channels.
Had ATI not did the experiments and wrote that paper, who knows how much longer it would have been before someone figured it out.
I think both methods should be supported. Especially since they're so cheap to implement, and very similar, it just makes sense.
akira888 said:
You guys continually miss the point. It's not simply "DXT5 compression". You're acting like what's being suggested is that developers simply compress normal maps like they were color textures with existing tools/APIs. In others words, do nothing different than they had been used to doing.
But you have to utilize it in a new way that many developers simply didn't investigate. You have to store your normals as tangent space normals, place one of the components into the *alpha channel*, write a new DXTC compressor optimized for handling normals, add 2 pixel shader instructions to all of your shaders that use them (or 3-5 if you want to renormalize)
Before ATI's paper, most developers probably just thought they could get away with low-res uncompressed normal maps (e.g. "they were good enough") ATI's evangelization of hi-res normal maps has now made this a public issue, so there will be more pressure.
Even after DXTC was well available in the market, developers were still using crappy 256x256 textures for a few years. These things take time to percolate.
I wouldn't call it "open" either - I guess it's in some kind of hinterland (the only info on Normal Map compression I can find on their website is, ironically, with regard to DXT5 compression). As I see it, if 3DC compression is as good as ATI lead us to believe, then it should (and would) be adopted in the DirectX standard. I can't see why anyone can argue that would be a bad thing? Why would it?akira888 said:
The only thing I would advocate is that new compression techniques be discussed in an open public forum like ARB. I don't like the "behind closed doors" approach of DX, where it is not clear who's doing what, who's talking to whom, and what deals are being made.
I prefer open standards groups, especially those that have rules against patents and royalty charges. For example, the W3C stipulates that anything developed in the context of working groups or committees must be freely licensed.
But more than that, having a group that freely takes draft submissions for new standards and evaluates them, picking a mutually agreed upon optimal course of action, is better overall for the industry. For example, ARB put out a call for OpenGL2.0 features. Many IHVs responded. They considered 3dLabs GLSL, NVidia Cg, and even Stanford's RTSL before deciding by vote to go with 3dLabs. Afterwards, some changes were made to 3dLabs proposal to accomodate issues that other ARB members had.
I prefer open standards groups, especially those that have rules against patents and royalty charges. For example, the W3C stipulates that anything developed in the context of working groups or committees must be freely licensed.
But more than that, having a group that freely takes draft submissions for new standards and evaluates them, picking a mutually agreed upon optimal course of action, is better overall for the industry. For example, ARB put out a call for OpenGL2.0 features. Many IHVs responded. They considered 3dLabs GLSL, NVidia Cg, and even Stanford's RTSL before deciding by vote to go with 3dLabs. Afterwards, some changes were made to 3dLabs proposal to accomodate issues that other ARB members had.
Diplo said:
Its already available in DirectX.
Being available as a FOURCC code, and being *required* in the core by MS are two different things. An effectively infinite set of different texture formats can be supported by the FOURCC mechanism, but until Microsoft says "it's required", it's merely an IHV-specific extension, like any other IHV specific extension.
Are DXT formats a requirment?
They are official microsoft formats. There are lots of D3DFMT_* codes that hardware can optionally support, but they are all Microsoft official in the core, not IHV specific extensions.
Don't try to tell me there isn't a difference between Microsoft officially sanctioning a format and an IHV extension.
Don't try to tell me there isn't a difference between Microsoft officially sanctioning a format and an IHV extension.
[url=http://msdn.microsoft.com/library/en-us/directx9_c/directx/graphics/reference/d3d/enums/d3dformat.asp?frame=true said:
DXT formats are still just fourcc formats in the DX spec, just like 3Dc is. They're just so commonly used that microsoft went a head and did a #define D3DFMT_DXTn = MAKEFOURCC('D','X','T',n), and sanctioned them in the runtime (such that the reference rasterizer correctly supports said formats)
Similar threads
