냐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� +-+-+-+-+-+-+-+-+ 栢栢栢栢昉께께굇굇굅같�|O|u|t|b|r|e|a|k|臍같굇굇굉께께栢栢栢� +-+-+-+-+-+-+-+-+ Issue #3 - Page 2 of 12 냐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� GeForce3 Refer to my earlier article on DirectX8 for clarity of performance issues. Graphics hardware has really matured over the past couple of years. The performance capabilities of today's 3D accelerators lets game developers truly display thier artistic skills. The emphasis is on realism and manufacturers such as nVidia have been consistently churning out the hottest gaming hardware to bring these effects to life. nVidia products usually score both in terms of sheer performance and thier use of innovative technologies encapsulated within as is the case with the GeForce3, which is also the only fully DirectX 8 compliant card around. Lets have a look at some of its features. The nfiniteFX engine One of the major enhancements in GeForce's technology is the ability to allow game developers to specifically program certain parts of this graphics processor allowing it to deliver custom special effects. These technologies are collective referred to by nVidia as the nfiniteFX engine. Two important components of this new architecture are the Programmable Vertex Shader and the Programmable Pixel Shader. Programmable Vertex Shader In any 3D model, the vertex is the most basic element. This is the point at which the polygons in a 3D model meet, so any changes in the position and other characteristics of a vertex can change the appearance of the 3D model. Until now, vertices could be assigened to perform a very limited number of operations in the 3D application. With the GeForce3, there are a number of custom special effects, such as morphing, animation, deformation and environmental effects, which can be applied to vertices. This gives programmers a high amount of flexibility in what can be done to the 3D objects in a game. Effects such as fog, lens effects and water refraction are now possible. Also, effects such as wrinkles and stretch marks can be created on the characters in games, objects can morph into other elements, and mechanical processes such as gear trains can be simulated accurately in real time. This means more natural animations for next generation game characters. Programmable Pixel Shader Like the programmable Vertex Shader, this feature allows application developers to take one step further towards creating life-like 3D scenes and effects. This technology has been around since the GeForce2 was launched. It was then known as the nVidia shading Rasteriser or NSR. Using it, effects such as reflections, bumps, fog and textures could be applied at the pixel-level in a scene. The GeForce3 processor takes this to a much higher level by allowing upto eight texture operations in a single pass. This is coupled with the ability to handle four textures at a time. The GeForce3 can combine these operations to produce breathtaking special effects such as reflective bump mapping and realistic textures with environmental effects such as fog and mist. Since these effects are processed in parallel, it doesnt affect the frame rates in games, but it certainly makes games look a whole lot better! Lighspeed Memory Architecture nVidia graphic cards have always been notorious for problems with thier memory bandwidth. As 3D processors keep getting faster, even tho exotic memory like DDR is used, performance is hit at higher resolutions because of the graphic cards memory arhitecture. Until now, the easiest way to counter this problem was to add more memory and pump up the memory's frequency. However with the GeForce2 Ultra's memory clocking at 460 MHz with bandwidths of upto 7.3 GBps something else had to be done to alleviate the memory bottleneck. So the GeForce3 uses a couple of innovative approaches to access and use its onboard memory. Splines: In the past, graphic applications used to send raw triangle data to the graphics card, which in turn would process this data and churn it out into displayable graphics. Nowadays the complexity of the 3D elements in games reaches high levels and there times when individual 3D elements are often composed of hundreds of thousands of polygons! So instead of relying on raw triangle data to define complex 3D surfaces, they are nwo defined by a number of control points, also known as splines. This results in significantly lower load on the memory bandwidth and greater efficiency is obtained, especially when working with complex 3D elements. Crossbar memory controller: This is another enhancement to the memory architecture. The GeForce3 uses four memory controllers that interact with each other and the GPU. therefore smaller chunks of information can be accessed and processed from many parts of the memory simultaneously making the entire operation faster. HRAA (High Resolution Anti-Aliasing) This is another feature that was incorporated to a lesser extent into the GeForce2 as well. Aliasing is the appearance of undesirable jagged edges in a 3D model when there are not enough pixels to represent it. The GeForce3 uses a process called Quicunx Sampling to minimise these 'jaggies' by interpolating pixels between the jagged edges. It does so by internally increasing the resolution of the game and then scaling it down to the required resolution. This feature is hardwired into the GeForce3; enabling HRAA to make edges and lines in 3D games appear smooth. Lossles Z-compression: One of the most important components of the 3D processor is the Z-buffer. This is the area where the 3D processor stores depth information about a scene. The graphics processor uses this information to know which 3D objects lie in the foreground and which are in the background. Since this information is extensive in typical 3D games, it takes up a lot of memory bandwidth, GeForce3 implements a form of lossless compression of this Z-buffer data and therefore allows the memory to be used more efficiently. Z-Occlusion Culling: When a 3D object is hidden by another object in the foreground, there is in fact no need to draw the hidden part because its not visible to the viewer. If this approach is used, the information needed for drawing this object does not have to be passed to the graphics processor. The GeForce3 uses this technique of only processing graphical information that will eventually be drawn in the scene. This results in a significant increase in the efficiency of the graphic processors output. Conclusion All said and done, the GeForce coupled with DirectX8 is greatly reducing the dividing line between games and movies. But are there really enough softwares to truly exploit all the features of the expensive GeForce3?? How many people have had enough of the GeForce3? To add to this dilemma, nVidia announced its latest chip on February 5, 2002 The GeForce4. The GPU code-named NV25 is closer to the processor that nVidia made for the Microsoft Xbox. A desktop version of the chip, code-named NV17, which was avaliable only in the mobile version will also accompany the NV25. It is definitely something to look forward to. The release date was pushed ahead, not surprising, considering that the GF4 is a lot more powerul than the GF3. It supposedly has six pixel pipes as compared to GF4's four pipelines. Other goodies avaliable in the NV25 are higher clock speeds, faster memory interfaces and improved antialiasing capabilities. Still in love with your GeForce3? A pity that you have to upgrade it even before you could even expolit it fully! - Strykar