A new graphicsengine from ATI called Charisma Engine makes a really "charismatic" impression. Just take a look at the list of itsfeatures.
by FastSite
03/09/2000 | 12:00 AM
All the information presented in the current article was obtained absolutely legally. We didn't break any laws and NDAs.This article contains some preliminary product features, which can be changed without giving any prior notice. All the infobelow should be regarded as our personal subjective opinion.<%BANNER[article]%>
It is not surprising for anyone that one of the today's key notions is "photorealistic 3D graphics", which implies that newcomputer games will look like a good Hollywood movie. Namely, it means that the game characters' facial expression will reflecttheir emotions and mood, the landscapes will include realistically rich greenery, the characters will be moving smoothly andnaturally, all scene objects will be correctly shaped and will cast natural shadows. Besides, the scene objects will be modeledat the high level of detail no matter if they are in the fore- or in the background, light reflections and environment will beconstructed as close to reality as possible, the shadows cast by all the light sources in the scene will change dynamically, thescenes will have lifelike sky, fog and clouds, and more, and more. Moreover, everything will change interactively depending on theplayer's actions in real time. Sounds impressive, doesn't it? Well, today it is more likely to seem a fairy-tale than a realstory. Any game developer as well as an ordinary gamer will definitely say that there is no appropriate hardware in the marketright now, able to make this dream a reality. Even the coolest today's CPUs won't be powerful enough to model such complicated3D graphics. Nevertheless, 3D graphics processor developers don't waste their precious time. Do you remember GeForce 256 fromNVIDIA - the first graphics processor with hardware T&L intended for the mass market? Most game developers were very enthusiastic aboutthe new features provided by HW T&L. However, we don't suffer from the bunch of games making use of all HW T&L advantages, but wehave to hand it to NVIDIA, they became the driving-force of the progress. Today there are quite many retail graphics cards witha GPU on board, and HW T&L pushes a lot of game developers towards significant improvement of gaming AI (we are fed up withstupid narrow-minded monsters, we believe) and makes them pay more attention to the game story and real physical lawsrealization (they don't need to develop their own software engine transforming polygon vertex coordinates). In general,everything happens as it actually should: the improvement of the GPU hardware potential allow game developers to devotemore time and effort to 3D graphics quality, the story, action, etc. They really get more time for creative work. That'swhy the architecture of most new graphics processors appears as complex as that of the CPU at least.
Each graphics processor is based on certain architecture and a number of technological solutions. ATI Company did itsbest to get ready for conquering the lead in the mainstream market of gaming solutions. So, very soon we will get acquaintedwith a new graphics processor aka under the codename Rage6. The heart of the new GPU will be made of Charisma rendering enginedesigned by ATI engineers. Rage6 chip will use new rendering architecture called Pixel tapestry also developed in ATI researchlabs.
Here we would like to comment on two things:
So, let's take a closer look at what Rage6 GPU actually is:
And now we suggest discussing a few peculiarities of a new chip and ATI's solutions in greater detail.
A new graphicsengine from ATI called Charisma Engine makes a really "charismatic" impression. Just take a look at the list of itsfeatures.
Hardware Vertex Lighting Setup. The Setup required to store the information on light sources and the to determine theiramount and brightness of light directed at each particular scene element.
Then the polygon vertex coordinates are transformed in hardware from the 3D coordinates of the modeled scene into 2D coordinates of themonitor screen. This operation should be executed because you can't use 3D coordinates to model an object in 2D space. Coordinatesare transformed in 2 stages. First, the initial 3D objects coordinates (called "object space coordinates") are transformed into thecoordinates of those objects displayed on the screen after rendering (called "view space coordinates"). Each pixel of the view spacecan be lit according to the info accumulated when the HW Lighting unit was working.
However, some other special operations can be executed, before the second stage of vertex coordinates transformation actually begins.
Vertex Skinning. This texturing method is based on skeletal animation. In fact, this strange explanation implies vertex blending.Vertex Skinning is used to provide smooth and natural picture when two textures meet, especially if the object is moving. Vertexblending is carried out with Skinning Matrices for Vertex Blending.
D3D supports 2- and 3-matrix vertex blending. In most cases 2-matrix blending is more than enough. However, when a neck and twoarms are jointed to the body, 4-matrix blending will be more preferable, and in case of a body and two legs - 3-matrix one.Nevertheless, you will be able to notice the quality improvement only if you attentively examine a full-screen man. GeForce 256,for instance, supports only 2-matrix blending, while Rage6 will also have 3- and 4-matrix blending. If you are looking for someillustrations of 2-matrix blending, take all the living creatures from DMZG demo, which makes full use of this effect.
Keyframe Interpolation. This technology allows changing the object's appearance on the level of hardware by setting only theinitial and the end images: all the transitional images are created automatically. In other words, Charisma Engine inserts therequired amount of interpolated frames between the key ones, which changes the character's facial expression in a very lifelikeway. Take a look:
After the special effects have been already applied, Clipping starts. Clipping means the deletion of the scene parts alongfour user defined surfaces (viewing frustrum + user defined clip planes). All the work is made on the polygon level,and not on the object one. This particular deletion method is the most frequently occurring in practice, because itallows to significantly reduce the memory bus workload during further rendering.
Now the geometric unit and the vertex lighting unit are more often referred to as HW TCL, which stresses the hardwareclipping, although the former abbreviation "HW T&L" is not in the least worse. This operation is also carried out on thehardware level and helps to avoid wasting time and effort during final scene rendering.
After the special effects comes hardware Clipping. This operation is also executed on the hardware level. It serves to avoid uselessmanipulations during the final scene rendering.
Now we are going to return to the second stage of vertex coordinates transformation, mentioned above. The coordinates aretransformed taking into account the user's view angle, the so-called correct viewpoint in the eye space. The coordinates aretransformed into 2D (screen coordinates), which corrects the distances. This operation is called perspective correction ortransformation.
After that, polygon vertexes form polygons and undergo texturing. And then the ready pixels with a certain color and certaincoordinates are displayed on the screen.
No doubt, TCL, Vertex Skinning and Keyframe Interpolation require very high computing capacity. Rage6 has to complete the taskseasily without overloading the system CPU. It will allow game developers to use more polygons for tessellation, which will definitelytell on the image quality and make, for instance, monsters more lifelike.
So, game developers will be able to use the uninvolved CPU resources for AI calculations, for example. Just for a clearer picture:the games without HW TCL usually have about 5% of the CPU resources for AI, while new games supporting HW TCL will be able to boast30-40%, which will make your virtual enemies much more inventive and cunning. Besides, the freed CPU resources can be also utilizedto calculate complex physical effects, such as gravitation and atmosphere phenomena. We should also say that if a game supports HWTCL but does not utilize the available CPU resources, you will still get very fast graphics of very high quality, although you won'tnotice any other improvements.
Besides Charisma Engine, Rage6 has another important feature - Pixel Tapestry (pixel rendering architecture). Its key peculiarityis three texture units in the rendering pipeline. If you remember, GeForce 256 and VSA-100 feature only one texture unit, Savage2000and NV15/NV11 - two texture units in each pipeline. Of course, Savage2000 is allegedly capable of laying four textures per pixel,but no one has ever tried it in real games for now.
Rage6 possesses two rendering pipelines with three texturing units each, which allows:
Besides, Pixel Tapestry provides hardware support of:
You know that the more textures are laid over one pixel peer clock, the more realistic and detailed turns the finalimage without sacrificing the general performance. Textures are used for laying lightmaps, shadow maps, specular maps,gloss maps, bump maps, etc.
We should also point out EMBM support, which is now featured only by G400 chips from Matrox but is already applied in afew tens of games.
A very important feature of Rage6 is Cubic Environment Mapping supported in D3D 7.0. This method is viewpoint independent,but uses 6 textures instead. Besides that, Rage6 also supports Spherical method (1 texture, viewpoint dependent) andDual-Paraboloid method (2 textures, viewpoint dependent).
There is a special technology used to make a scene look more lifelike and to create scene depth - Shadow mapping.Rage6 supports a special Priority Buffer, which stores the info about the closeness of the objects to a light source.So, it assigns a weight coefficient (or priority) to each scene element. If the object's location relative to theviewpoint changes, the coefficient also changes. The combination of this Priority Buffer and Projective Textures givesvery realistic shadows. The shadows are created in the following way. The light source is taken for a viewpoint and theentire scene is then rendered depending on the weight coefficients from the Priority Buffer: the closest object is thefirst to be rendered because it has the highest priority coefficient. Shadow texture is formed by rendering the silhouetteof each object. Then the ready texture is projected and we get shadows. This method is much easier and more efficient thanthe traditional stencil buffer.
Rage6 supports Range-Based Fog, i.e. the fog density depends on the distance between the object and the player.
One of the newest functional features of Rage6 is 3D textures support. 3D Texture is a 3D pixel array and not a 2Done as usual. This is a totally new thing for OpenGL 1.2. Some time ago Permedia3 from 3Dlabs was the only one supportingthe textures of this type. 3D textures are perfect for voxels and spherical textures used to create dynamic globs of light. 3D texturessimplify the use of deformable geometry. Parts of objects can be chipped or cut away without having to update textures.
Well, let's return to Rage6 chip, especially since it is not a dream any more but a reality. Note that starting from A0 chiprevision (alpha-version) the chip is absolutely ready for normal operation and is deprived of serious hardware defects. Toillustrate this statement and to prove that the graphics card built on it, the drivers and BIOS also work well we would liketo draw your attention to the fact that it works perfectly in Quake3. This success attests high quality of the design andwell-arranged work of ATI's engineering team, although Rage6 architecture is very complicated. By the way, if you rememberRage128 chip didn't please us that much: from the very beginning it had a few design defects, which improvement consumed alot of precious time and effort. Of course, the test versions of Rage6 based cards work at lower frequencies that is why itis still too early to discuss the performance of the future card. However, the chip architecture gives us to understand thatit will be cooler than GeForce 256 not by a few percents but in several times. The main competitors of Rage6 will certainly beNV11/NV15 based cards, and this competition is going to be really cut-throat in terms of performance and features versatility.
You probably wonder why we are so sure? Take, for instance, today's main competitors of ATI Rage6 - GeForce 256 basedcards. To make it simpler, GeForce's 256bit core is made up due to 4 64bit units working in parallel and connected to oneand the same internal bus. As a result, the mutual electromagnetic influence (and to be more exact - mutual signal fronts'influence) is so high that the core working frequency and performance get significantly lower. To combat this unpleasant effect NVIDIA had to cutdown the core frequency of its GeForce 256 to 120MHz so that it worked more stable. Even the use of 0.22 microntechnology doesn't affect so greatly the general performance as the architectural peculiarities. Of course, this is arelatively primitive explanation: we didn't touch upon a number of other factors. However, it seems to be enough to giveyou an idea of what we were talking about.
Rage6 can boast a much finer realization. Primitively speaking, instead of four 64bit units there are two 128bit ones connected in parallelto the internal bus via special buffer, which positively tells on the performance. Therefore, even when working at equalclock frequencies, Rage6 will perform about 1.5 times faster than GeForce 256. Nevertheless, we think that Rage6 owes thisperformance to internal chip architecture intended for higher clock frequencies. They can be achieved due to the use of arendering pipeline requiring a greater amount of simpler stages (compared to GeForce 256). Since 3D graphics algorithms aremore regular and a large number of pipeline stages doesn't cause any serious delays in case pipeline restarting is needed.We will be able to check if our expectations aretrue as soon as we get a working Rage6 based card. And for the time being all the info is just the info: no proven facts.
Rage6 based graphics cards are expected to appear in retail in Q2, maybe in May. A card on this chipset with 32MB DDR SGRAM willsupposedly cost around $250. The chip will probably work at 250MHz and the memory - at 200MHz.
If you don't have AGP interface in your PC (e.g. if you work on i810 based mainboard), you are probably curious to knowsomething about the PCI version of Rage6 based card. Here we can't now give an exact answer. From the technical point ofview, nothing should prevent ATI from working out a PCI version as well. However, it will still greatly depend on the demandand marketing.
Of course, we can also look forward to graphics cards on two Rage6 chips (or even more) with MAXX technology working inparallel. Although these cards won't become a mass product, hardcore gamers should be really happy with them.
If we take a look at the list of hardware requirements for 3D graphics accelerators of the latest generation introduced byMicrosoft, we will see the following:
You can easily notice that Rage6 fully complies with Microsoft requirements. ATI has every chance to become if not the mainpartner then at least one of the main Microsoft partners in supplying the X-Box package with a graphics subsystem. Moreover,since there is still a lot of time left until the launching of X-Box and Rage6 looks very attractive with NV15 in the background.
So, we can state that ATI Rage6 is undoubtedly a second generation GPU featuring a lot of architectural innovations and hugepotential.
The most important thing now is to launch the cards on Rage6 in Q2 2000 and not just to provide a few pieces somewhere but toarrange mass supplies at a reasonable price. And in the meanwhile we can just patiently wait aside and hope that ATI has alreadylearned its lesson and obtained enough experience during Rage128 Pro and RAGE FURY MAXX sales.