DirectX9 Preview
25. March 2002/of aths & Leonidas /page 1 of 1
Microsofts DirectX covers a Sammelsurium of routines, in order to standardized and obtain fast access to hardware under Windows. In things diagram went it at that time with "WinG" loosely. This was an extension for Windows 3,1, in order to smuggle through 2D-Ausgaben at the slow Windows GUI. Under Windows 95 then in an extended version the DirectDraw , which was later then combined with other things into
DirectX, was called. Us interested of it naturally in particular the interface in the 3D-Hardwarebeschleunigung, speak
Direct3D.
The most standing out characteristic of
DirectX7 was as well known the support for hardware T&L acceleration
(which it for OpenGL however already longer gave). At that time nVidia licenced appropriate DirectX procedure from Microsoft, and brought well-known GeForce chip the first T&L accelerator on the Consumer market. For
DirectX8 and its predominant adjustment on programmable Vertex and pixels Shader engaged themselves beside nVidia above all still 3dfx - and Matrox. We want to regard now some important innovations of the coming
DirectX9 in the following.
Matrox seems to play here a role of the pioneer with DirectX9. For months Matrox has spoken about
DISPLACEMENT Mapping : This technology one can understand oneself only once as advancement of Bump Mapping. While Bump Mapping rough surfaces however only simulated - the polygone remain smooth - "are indented" these by DISPLACEMENT Mapping actually. A large advantage resulting from the fact is that DISPLACEMENT Mapping for wide geometry can be used - up to whole landscapes, while Bump Mapping is suitable for small details particularly.
Now curved surfaces can be reached also with Curved Surfaces. DISPLACEMENT Mapping permitted however on simpler way of more details to represent and the procedure works without not fitting connection points
(which in all other respects also a reason for it it were, why John Carmack Curved Surfaces for "overrated" explained). Further advantages of DISPLACEMENT Mapping are the high compression rates
(geometry can be compressed thereby approximately up to 40:1!), with which the AGP bus is relieved. In addition a level OF detail system is automatic with integrated.
It is obvious that conventional 3D-Hardware cannot naturally be capable of DISPLACEMENT Mapping. Because around DISPLACEMENT Mapping meaningfully to begin, must support the
GPU adaptive Tesselation (dismantling of the object data in polygone ) and not only simple transformation. Past GPUs must process and be able to do all corner points it neither omit, nor their own produce itself. If a diagram map transfers now the Tesselation, thus the dismantling to triangles, relieved that not only the CCU and the AGP bus, but overcomes also this past restriction. Can DISPLACEMENT map internally like textures be treated - inclusive MIP Mapping .
Picture in attachment
DISPLACEMENT Mapping: From a simple surface fissured mountain country shank charms (click to magnify)
In the future
Shader with DirectX9 seems to indicate it will compete with the clay/tone of ATi. Both Vertex and pixel Shader will develop ever more toward "genuine" programmability - bypasses and loops become possible. The pixels Shader version 2.0, which will be introduced with DirectX9, obviously rather pixel Shader realization addresses itself on nVidia's 1.1er to 1.ér pixel Shader realization out to ATi's 1.êr (3dfx ' last and unpublished chip "RAM PAGE" supported version 1.0) .
Particularly interesting are here the data formats. With DirectX9 the possibility of the 64-Bit-Renderings introduction will hold - whereby in addition not only fixed comma is possible -, but also for floating point representation. The advantage of it is that the relative error is distributed evenly with all pixel operations on all brightness. While with DirectX8 the 12 bits accuracy per colorchannel are only one recommendation, DirectX9 nevertheless requires a mindeste internal accuracy of 10 bits per colorchannel.
In order to summarize again briefly, which the most important components of DirectX9 will be:
- Vertex Shader 2.0
- Pixel Shader 2.0
- adaptive Tesselation
- DISPLACEMENT Mapping
- at least 10 bits internal accuracy per pixel colorchannel[/list:u] In what respect these features are used however as soon as possible by the game programmers, is naturally now still most uncertain. DISPLACEMENT Mapping is this feature, which might be most lucrative probably for the game programmers, can a first-class diagram also quite little expenditure be reached nevertheless. Also the Matrox its DISPLACEMENT Mapping already since longer the game programmers tastily to make tried, might proving with dessem soon-possible employment than useful.
The drawback the medal is naturally again and again the sufficient availability of DirectX9 hardware - and without these runs with the allowance of the play programmers to new features little according to experience. The old "hen egg problem" will meet the DirectX9 feature & hardware of our opinion after in particularly hard measure: Because the game programmers turn at present straight only once DirectX8 and its utilization. DirectX9 is from this view a good continuation and extension existing, contrary to DirectX8 however no necessity.
To that extent we think at present that it will have DirectX9 still more with difficulty than DirectX8 to become generally accepted with the game programmers. If we must set approximately 1.5 years with DirectX8, before there are appropriate games in mass, this will rather longer last with DirectX9-Spielen therefore in no case more briefly, but. Ergo: If DirectX9 this summer/autumn appears, it becomes probably hardly with DirectX9-Spielen in mass something the year ago 2004.
To the topic of the DirectX9 hardware ansich: It is unmistakable that for DirectX9 above all ATi and Matrox engage themselves strongly. That permits the assumption that both want to come with appropriate hardware in time and particularly as first on the market. Matrox seems to want to tie here to old times and would like again once a role of the pioneer to probably take over, however at present still are more exact plans of Matrox in the dark one.
With Ati is against it well-known that chip in this autumn (some talk about this summer) will begin their DirectX9-Chip R300, first data gives it in an older ATi Roadmap . Too nVidia's DirectX9-Chip NV30 are present themselves against it as with Matrox so far hardly data, only with release are quite safe one (so far) due to the usual nVidia Produktzyklus: Autumn this yearly.
Whether it will come to who will bring the first DirectX9-Chip on the market and when DirectX9 will now begin - we to become it in the run this yearly to experience to be allowed: -). Zumindestens appears it at present in such a way, than that for lack of DirectX9-Games nobody will have to mothball its DirectX8-Hardware with the appearance of the first DirectX9-Chips.
Download on the left of: Microsoft DirectX9 presentation + Matrox DISPLACEMENT Mapping presentation (both PPT)
OGL drajveri za Matrox kartice su oduvek bili loši, a ovi za G400 su "samo" spori, dok stabilnost nije dovedena u pitanje. Već sam objašnjavao da je to verovatno nameran strateški potez u nekakvom dilu sa Bilijem. Koliko osnova ima za ovakve sumnje možda najbolje govori podatak o njihovoj vezi ( M&M ) i zajedničkom radu na razvoju DX9 API-a.
