Glaze 3D |
||||
Glaze3D™ is the fastest yet consumer-level integrated 2D/3D graphics processor core. Delivering 400 million dual textured, illuminated, fogged, alpha blended, anti-aliased, Z buffer pixels per second, the Glaze3D™ sets a new standard on 3D graphics performance.
With a straightforward, easy-to-implement rendering pipeline built using custom-designed algorithms the Glaze3D™ graphics processor core architecture increases chip vendor's profit margin by providing the best possible performance with a minimum amount of silicon. Highlights
Simply put, Glaze3D™ is four times faster than any consumer-level 3D-graphics chip available today. The fastest chips available today can render 90 million dual textured pixel per second. With 400 million rendered pixels per second, Glaze3D™ leaves others standing in the dust. The Glaze3D™ has been designed for screaming performance in games and business applications, not just in benchmarks. The architecture allows features to be turned on without compromising on speed.
Dual texture pixel fillrate has doubled every 18 months. With 400 MPIX/s dual texture fillrate the Glaze3D™ will be a tough competitor in the year 2000 and its scalability allows future growth to maintain its performance lead. Glaze3D™ has been designed to render translucent surfaces with full 400 million rendered, dual-textured, alpha blended pixels per second. Most of the required pixel fillrate in today's computer games comes from translucent surfaces such as smoke, lighting effects, clouds and explosions - these can easily double or triple the required pixel fillrate. The integrated 2D-acceleration of Glaze3D™ is fully Microsoft Windows compatible, the fills and bit-block transfers operate at the speed of 400 million pixels per second. An integrated programmable floating point accurate triangle setup engine can handle a constant feed of 3 million fully-featured triangles per second. Glaze3D™ is being performance tuned on our behavioral software simulator of the core architecture, pictured here. We use test data from games such as id Software Quake and Quake II and from benchmarks like the Viewperf so we can test with real-world application data. The model gives us cycle-accurate information about the rendering time and possible bottlenecks in the architecture. |
This image shows a frame from id Software Quake II. It's a relatively complex frame with 4003 triangles and depth complexity of 1.93, meaning that every pixel on the screen is rendered twice for every frame. This particular data has been generated using multitexture rendering support, so it doesn't have separate lightmaps. Glaze3D™ features Glaze3D™ has been designed to be compatible with the DirectX® and OpenGL® APIs. The architecture supports perspective correct diffuse and specular shading, perspective correct per-pixel fogging, fog table, edge anti-aliasing, stencil operations, Z buffering and all alpha blending modes. The Glaze3D™ has a full 32-bit true color rendering pipeline, providing maximum accuracy and image quality. True color framebuffer and 32-bit floating point Z buffer add up to the visual rendering quality, eliminating the artifacts common in 16-bit rendering architectures. Four simultaneous textures are supported in a single rendering pass with trilinear mip-mapping. Most architectures support only one or at most two textures in a single pass, forcing game developers to use multiple rendering passes, which increases the geometry processing requirements and decreases rendering quality. Glaze3D™ also includes the bump mapping technique now included in Direct3D® 6.0 by Microsoft. Bilinearly interpolated bump map can be combined with a surface texture, this all is fully supported in hardware in a single rendering pass. Bump mapping allows rich surface detailing to be done without adding polygons to the surface geometry. Glaze3D™ architecture Glaze3D™ has a tried and tested rendering architecture, no tricks, just pure screaming performance. The architecture has been balanced between speed, features and silicon size to offer a cost-effective single-chip solution for thirsty gamers and power-users. The architecture is scalable from 200 to 800 million rendered pixels per second. This supports the diverse requirements of the graphics accelerator market by allowing different versions of the core to be used on products targeted at different markets. The speed of the triangle setup engine is scalable from 3 million triangles upwards and a full geometry processor design is also available for the core, to either supplement the triangle setup engine, or to replace it. The memory interface of the Glaze3D™ is flexible, allowing different types of memory to be used, based on the bandwidth requirement of the Glaze3D™ configuration. The current 6.4 GB/s bandwidth, required for true color rendering of 400 million pixels per second is achieved using Rambus DirectRDRAM memory. The memory interface is capable of utilizing SDRAM, SGRAM, RDRAM and SLDRAM memory types, among others. Drivers It's nowadays too common to ship new graphics accelerator boards with unfinished drivers. Bitboys Oy has solved this problem with the new revolutionary PCIBuilder™ development system, which has allowed us to work on the drivers well before the hardware is available. Windows NT driver for the Glaze3D™ was developed early in the architecture design phase and work has already begun on Windows 95®, Direct3D® and OpenGL® drivers. Reference drivers for the Glaze3D™ architecture will be available before the product ships, allowing the hardware to be shipped with solid, performance tuned drivers. |
|||
Last Updated May 26th, 1998 |