NVIDIA GeForce 256 Introduction

• 15M triangles/sec With sustained DMA. All transformed, clipped and lit.
• 4 Parallel rendering pipelines giving four pixels per clock. Clocked at 120 Mhz at launch but likely to come out in faster flavours next year.
• 480M pixels/sec fill rate - 32 texture samples per clock, free 8 sample anisotropic texture filtering.
• 480M texels/sec texturing rate.
• 8 lights in hardware.
• 350 MHz RAMDAC.
• Support nearly all features DirectX7 and OpenGL, transform & lighting, cubic environment mapping, projective textures and texture compression.
• 4x AGP performance with Fast Writes, which enables the CPU to send data directly to the GPU bypassing system RAM, thereby increasing overall performance and freeing the system memory bus for other functions.
• 256 bit rendering engine for better high resolution performance.
• 5 horizontal, 3 vertical taps.
• 8:1 up/down scaling.
• Independent hue, saturation and brightness controls in hardware.
• Highest quality HDTV (High Definition Television) video playback.
• High Precision HDTV video overlay.
• HDTV motion compensation.
• High bandwidth HDTV class video I/O.
• 16 bit video port.
• Full frame rate DVD up to 1080i resolution.
• Sub-pixel accuracy to 1/16 pixel.

Hardware Transforms

Probably the primary new feature. If you look at the history of graphic cards in the PC arena, more and more features have left the CPU. This is the latest move in this trend. As you all know the currently favoured method of drawing 3D images on screen is to describe the entire scene in terms of polygons, many sided shapes, mostly triangles.

These polygons are defined by their corners or vertices. Each of these is described by three numerical values which denote a point in the width, height and depth of space. The catch is that the viewer has his own position and direction of view. This is accounted for by assuming the screen is an imaginary flat area (plane) onto which all this vertices are projected onto. This process is called transformation and those of you who can remember their matrix maths will probably understand the concept quite well.

What you end up doing is converting the (x,y,z) cordinates of models into the proper (x,y) cordinates on the computer display. Considering that a scene can be involve quite a bit of polygons each of which have many vertices this works out to quite a workload for the CPU. This has in general kept the polygon count of games down. Instead game developers have tried to fool the eye into seeing detail by having relatively boxy models draped in detailed textures.

A good example is racing games, take Need For Speed for instance, all the track side trees are just flat surfaces of maybe two triangles with a image of a tree draped over it, use of transparent areas in the texture makes this possible. Now the ideal situation of course would be to model the tree down to more details but this would have brought the CPU to it's knees. Even today's high-end CPU's like the Athlon can't manage this and still allow for the game logic. Judging by the in-game screen-shots of some of the games put on display it seems the GeForce can do this! Let's take a look at some pictures now, all courtesy of The Whole Experience.

      
You can view the rest of the screen shots at Experience 3D

Before we leave the topic there is one last thing I have to touch on. If your into 3D graphics you may have wondered about how different high-end professional 3D graphic cards are from consumer gaming cards.

Other than the very obvious price difference (which is quite staggering!) they always have very high geometry capabilities and relatively limited fill-rates. This is the reason that the Intergraph Realizm boards Id Software have on their workstations aren't really the best thing for playing Quake. The low fill-rates stem from the fact that the cards are usually rendering to a small view-port. But the high speed geometry capability is essential for detailed images. In fact high speed geometry acceleration is just as important for Hollywood quality graphics as the T-Buffer features. The GeForce differs from the professional cards, which have lower fill-rates than current gaming cards, by having higher fill-rates since it has to do full-screen rendering.

Next page: Vooddo 4 vs. GeForce 256