blastarr
Power Member
DirectX 10
"With all the talk of next-generation consoles—and some very impressive screenshots floating around the web—fans of PC games are naturally wondering whether these powerful new systems are going to "kill" PC games. In a word . . . no.
The upcoming Xbox 360 and PlayStation 3 are based on DirectX 9 technology, in the case of the 360 it's "DirectX 9 and then a little more." But DirectX 10 is a whole different animal. It's a major revision to the API, almost a complete rewrite that requires substantially different hardware than the stuff we've seen so far.
To start with, let's clear up a few naming misconceptions. Over the past year or more, the graphics "stuff" coming in Windows Vista has been referred to by many names. DirectX Next and Windows Graphics Foundation 2.0 are two of the most prominent. The names have been changing internally at Microsoft, and it seems that they've all but settled on actually calling it DirectX 10. Contrary to some reports, it will ship with Vista, along with DirectX 9.L, a version of DX9 altered to fit the new LDDM driver model used by the OS.
DirectX 10 started by fixing what was broken in the previous APIs, like some stability problems and small batch performance, and then removing old unnecessary parts of the API (like the fixed function transform and lighting calls). This served as the foundation for a graphics API that could radically change the way games look and really take PCs to that next quantum leap, even over next-generation consoles.
The new graphics API will have much more stringent requirements for graphics cards, with a very particular guaranteed feature set. There should be no more "cap bits" needed to determine if your graphics cards can perform certain functions. The behavior of DX10 cards will be strictly defined, so developers can get the expected output from their code with no tweaking necessary for the eccentricities of different graphics cards from different vendors.
It also requires several new features of the hardware. The first is a new "geometry shader" function, which operates not on single vertices like today's vertex shader units, but on entire primitives: dots, lines, lines with adjacent vertices, triangles, and triangles with adjacent vertices. The huge performance penalty imposed by too many state changes should be a thing of the past as well. Render states are grouped into five different objects that can be cached by the hardware, with up to 4096 state objects of each type cacheable at once. DX10 also introduces a common shader core between pixels and vertices. Granted, this does not mean that the hardware itself needs to have ALUs that operate on either pixels or vertices, just that the language and functions have been fused into a single shader set.
The net result of these things should be games with an absolutely unprecedented level of detail, including a dramatic increase in "clutter," or the hordes of random and different stuff that exists in the real world but not in games. Obviously, rendering quality will shoot way up, too, with improved masking functions for antialiasing. It will also mean better object sorting, the ability to algorithmically generate content entirely on the GPU, and ultimately memory virtualization in the LDDM driver model to reduce bandwidth costs and provide more granular access to graphics data.
Right now, it's all a bit too much to take in. Some of the specs are still in flux, and you need an unabridged programmer-to-English dictionary to even comprehend the scope of the changes and their ramifications. Suffice it to say: When DirectX 10 games hit us, they're going to be of a quality that next-gen consoles can't touch.
The Future's So Bright
It looks like Microsoft is really doing all it can to move gaming and 3D graphics on Windows forward with the new Vista OS. Bear in mind, we've only just heard about a few of the features. Most of the good stuff is actually going to premiere in Beta 2 and beyond. WinSAT, for example, is only a shadow of its eventual self in the current Beta 1 build. There are lots of features that Microsoft won't show or even talk about for months.
What's clear is that Microsoft takes games on the PC very seriously and intends to make a Windows Vista PC one of the most vibrant and attractive game platforms anywhere. The goal is to treat the launch of Vista like the launch of a major console, with gotta-have-it games optimized for the OS, stores devoting lots of shelf space and big displays to the hottest titles, and gamers lining up to get their hands on it. Microsoft still has a long way to go to convince gamers that the PC is the place to play, and even farther to get them to shell out money for an OS upgrade to Windows Vista to do it, but it's definitely on the right track and has plenty of time to go."
in Extremetech
"With all the talk of next-generation consoles—and some very impressive screenshots floating around the web—fans of PC games are naturally wondering whether these powerful new systems are going to "kill" PC games. In a word . . . no.
The upcoming Xbox 360 and PlayStation 3 are based on DirectX 9 technology, in the case of the 360 it's "DirectX 9 and then a little more." But DirectX 10 is a whole different animal. It's a major revision to the API, almost a complete rewrite that requires substantially different hardware than the stuff we've seen so far.
To start with, let's clear up a few naming misconceptions. Over the past year or more, the graphics "stuff" coming in Windows Vista has been referred to by many names. DirectX Next and Windows Graphics Foundation 2.0 are two of the most prominent. The names have been changing internally at Microsoft, and it seems that they've all but settled on actually calling it DirectX 10. Contrary to some reports, it will ship with Vista, along with DirectX 9.L, a version of DX9 altered to fit the new LDDM driver model used by the OS.
DirectX 10 started by fixing what was broken in the previous APIs, like some stability problems and small batch performance, and then removing old unnecessary parts of the API (like the fixed function transform and lighting calls). This served as the foundation for a graphics API that could radically change the way games look and really take PCs to that next quantum leap, even over next-generation consoles.
The new graphics API will have much more stringent requirements for graphics cards, with a very particular guaranteed feature set. There should be no more "cap bits" needed to determine if your graphics cards can perform certain functions. The behavior of DX10 cards will be strictly defined, so developers can get the expected output from their code with no tweaking necessary for the eccentricities of different graphics cards from different vendors.
It also requires several new features of the hardware. The first is a new "geometry shader" function, which operates not on single vertices like today's vertex shader units, but on entire primitives: dots, lines, lines with adjacent vertices, triangles, and triangles with adjacent vertices. The huge performance penalty imposed by too many state changes should be a thing of the past as well. Render states are grouped into five different objects that can be cached by the hardware, with up to 4096 state objects of each type cacheable at once. DX10 also introduces a common shader core between pixels and vertices. Granted, this does not mean that the hardware itself needs to have ALUs that operate on either pixels or vertices, just that the language and functions have been fused into a single shader set.
The net result of these things should be games with an absolutely unprecedented level of detail, including a dramatic increase in "clutter," or the hordes of random and different stuff that exists in the real world but not in games. Obviously, rendering quality will shoot way up, too, with improved masking functions for antialiasing. It will also mean better object sorting, the ability to algorithmically generate content entirely on the GPU, and ultimately memory virtualization in the LDDM driver model to reduce bandwidth costs and provide more granular access to graphics data.
Right now, it's all a bit too much to take in. Some of the specs are still in flux, and you need an unabridged programmer-to-English dictionary to even comprehend the scope of the changes and their ramifications. Suffice it to say: When DirectX 10 games hit us, they're going to be of a quality that next-gen consoles can't touch.
The Future's So Bright
It looks like Microsoft is really doing all it can to move gaming and 3D graphics on Windows forward with the new Vista OS. Bear in mind, we've only just heard about a few of the features. Most of the good stuff is actually going to premiere in Beta 2 and beyond. WinSAT, for example, is only a shadow of its eventual self in the current Beta 1 build. There are lots of features that Microsoft won't show or even talk about for months.
What's clear is that Microsoft takes games on the PC very seriously and intends to make a Windows Vista PC one of the most vibrant and attractive game platforms anywhere. The goal is to treat the launch of Vista like the launch of a major console, with gotta-have-it games optimized for the OS, stores devoting lots of shelf space and big displays to the hottest titles, and gamers lining up to get their hands on it. Microsoft still has a long way to go to convince gamers that the PC is the place to play, and even farther to get them to shell out money for an OS upgrade to Windows Vista to do it, but it's definitely on the right track and has plenty of time to go."
in Extremetech
