Currently, extension modules for VGA cards with 2(4) MB DRAM based on chipset S3 Virge (Data Expert product) are available. This module is connected on the VGA Feature Connector and the entire system operates in standard applications the same way as a standard VGA accelerator.

At present, software libraries using special HiFill features are under development.

Evaluation kit for software developers including PCI graphic card based on S3 Virge with 2MB DRAM + HiFill Accelerator module is available now.

PCI Graphic Card without and with HiFill accelerator instaled

ISA verion of HiFill VGA accelerator

Hardware requirements for running graphics on standard PCs with HiFill are dramatically lower. At the same time, performance results are impressive. These two factors will allow virtual reality and high quality games software to become widely available. The PC systems equipped with the HiFill Graphic Accelerator will perform as well as high performance Silicon Graphics workstations. The graphic accelerator HiFill works without delay and can be used on virtually any host system.

The main problem of graphic systems currently in use is the volume of data transferred between the screen and video RAM. The video bus creates a bottleneck for all the sophisticated graphical applications on standard PCs. The core of HiFill acceleration method is reduction of data transferred into video RAM. HiFill transfers only information about polygons, their color, gradient, and texture. Picture 1 shows a triangle with a gradual/diant fill on a screen. In a standard graphic processing to show and delete this image on a 1024x768 pixel screen with 256 colors per pixel would require 76.8 KB of data. In HiColor (64k colors/pixel) it would require 150 KB, and in TrueColor (16M colors/pixel) 300 KB. Picture 2 shows data transfer requirement for the same image using HiFill graphic acceleration (3KB).

Pic.1Pic.2

The speed depends on the level of detail of a particular image. The HiFill acceleration allows for better performance and higher speeds for images with higher complexity. This is particularly useful for applications such as high resolution screens and virtual reality helmets, where deeper immersion in the scene is possible.

System HiFill is fully open, even in comparison with systems based on Open GL, and accelerates images based on any rendering algorithm. Fractal structures allow for a high quality pictures, while efficient data formats allow for compressions higher than JPEG or MPEG. As opposed to the current graphic accelerators, which accelerate 2 to 4 times, HiFill is capable of accelerating 10 to 100 times, while maintaining the quality of graphic workstations. HiFill uses efficiently current technology since it demands increased activity from the microprocessor, while limiting the use of bus. Additionally, HiFill allows showing images in so called HiFill Extended Color Mode. This mode allows the picture quality close to the TrueColor regime (4 - 32M colors depending on implementation), while the host system maintains setting of 256 colors. That way a high quality image with high resolution can be displayed with limited demands on video RAM. For instance, for two pages with resolutions 1024x768 pixels 2 MB of VRAM is sufficient, while 800x600 requires 1 MB of VRAM only. Consequently, high resolution 3D and high frequency (60 frames/second) scenes can be displayed on standard hardware (Pentium chip with PCI bus).

Technically, HiFill allows for several integration levels. Currently, extension modules for VGA cards with 2 MB DRAM based on chipset S3 Trio64+ and Virge are available. This module connects to VGA Feature Connector and the entire system operates the same way a standard accelerator would. Furthermore, under preparation is an option of connecting HiFill into RAM DAC of graphic adapters, into VGA chip, or into graphic adapters with SRAM for Intel AGP standard.

Following Chart 1 and Picture 3 compare performance of graphics of SGI Indigo Maximum Impact with Pentium 90 PCI PC equipped with HiFill on a host card SVGA with 2 MB DRAM/S3 Trio 64+. For comparison, a standard Tmesh test (10x10, 5o pixels isosceles triangles shown for a second) and modified Tmesh FullScreen test, where size of the of the triangle is determined by the size of full screen. Last part of the chart shows the price comparison.

Silicon Graphics Indigo2 Maximum Impact / HiFill PC

HiFill - BREAKTHROUGH SPEED LIMIT

HiFill accelerate scanline algoritmus, because scanline renderers is quick (RenderWare) then OpenGL in Win95 and high quality scanline render's is available (radiozity, shaders - parametric texture, nurbs....with small models). Next algoritmus for games is horizontal cliping and X-buffering - right-left clipping and S-buffering, front to back writing. HiFill will use for two level texturing, wher texture may be more zoomed, you write only first repeating pixel in scanline and HiFill use for second level litlte patern of texture (that will use small texture in RAM and big level of detail on screen).

HiColor data formate

8pixel texture modulate runing colour - intenzity/2 and is zooming 1x;1,5x;2x;2.5x;3x;3.5x;4x

integration - ADD or SUB gradient intenzity +/-0;1;2;3;4;5;6;7 in TrueColor 8.8.8 to white or black

gradient -0 is interpolation texture - bits is signum of gradient

backcolor must be FFh, synchronization must be -V, +H and pallete in RAMDAC all set to black

(S3 Virge have non Intel convention WR to memory -big/little endian)

.

download demo + source code for WATCOM 10.5

DOWNLOAD old demo (VESA 2.0)

demo for Win95 - demo go as doublebuffering for one open windov

DOWNLOAD Win95 demo (DirectX)

Our product direct to Microsoft's Talisman project

Now we are working on TurboRey - fast scanline resterize engine for HiFill with 32bit Z buffering, alpha blending, fractal LOD texturing, radiosity shading, fractal objects and Nurbs and Metaballs (all calculated by rasterization to keep small model and fast 3D transformation for easy transfer on Web)

..................last actualization 1.1. - 1998