History of video processors: 1976 - 1995

image






The evolution of modern graphics processors began with the appearance in 1995 of the first 3D-cards, followed by the widespread adoption of 32-bit operating systems and inexpensive personal computers.



Prior to this, the graphics industry mainly consisted of more prosaic two-dimensional non-personal computer architectures. Graphics cards were most often known by the alphanumeric designations of the chips and were of enormous cost. 3D games and computer graphics gradually came together, arising from such diverse sources as arcade and console games, defense industry projects, robotics and space simulations, as well as medical imaging.



The early days of consumer 3D graphics were the Wild West of competing ideas: from ways to implement equipment to using various rendering techniques and application and data interfaces. The first graphic systems had a fixed function pipeline (FFP) and an architecture that followed a very strict processing path; at the same time, almost every 3D chip manufacturer had their own APIs.



3D graphics turned a rather dull PC industry into a show of light and magic, but it owes its existence to many generations of innovation. In this series of articles, we will take a closer look at the history of the GPU: the early days of consumer 3D graphics that revolutionized 3Dfx Voodoo, industry consolidation at the turn of the century, and modern GPGPUs.



1976 - 1995: the first days of consumer 3D graphics



The first real 3D graphics came with the advent of display controllers called video shifters or video address generators. They acted as an intermediary between the central processor and the display. The incoming data stream was converted into a sequence of output video bit data, such as brightness and color, as well as into vertical and horizontal composite synchronization, which stored a row of pixels and synchronized each subsequent line and the interval of the quenching pulse (the time between the completion of one raster line and the start of a new one).



A variety of designs of such devices appeared in the second half of the 1970s and laid the foundations of the 3D graphics known to us.









Atari 2600 Released in September 1977



For example, the Pixie video chip (CDP1861), released by RCA in 1976, was able to output NTSC-compatible video with a resolution of 62x128 or 64x32 for the unpopular RCA Studio II console.



One year later, the video chip was followed by the Television Interface Adapter (TIA) 1A, which was integrated into the Atari 2600 to generate screen images, sound effects, and read information from input controllers. TIA was led by Jay Miner, who later designed chips for the Commodore Amiga computer.



In 1978, Motorola announced the release of the MC6845 video address generator. It became the basis for the Monochrome / Color Display Adapter (MDA / CDA) cards for the IBM PC, released in 1981. This generator performed the same functions on the Apple II computer. Later that year, Motorola added the MC6847 video display generator, which began to be used in many first-generation personal computers, including the Tandy TRS-80.









Monochrome Display Adapter for IBM PC



A similar VIC device, manufactured by Commodore MOS Tech, provided graphics output to Commodore 1980-83 home computers.



Next November, the ANTIC (Alphanumeric Television Interface Interface) controller and CTIA / GTIA (Color / Graphics Television Interface Adapter) coprocessor debuted in Atari 400 computers. ANTIC, using direct memory access (DMA), processed 2D image commands. Like most video coprocessors, it could generate game field graphics (backgrounds, splash screens, glasses display), and CTIA generated colors and moving objects. Yamaha and Texas Instruments also supplied similar integrated circuits to manufacturers of the first home computers.



The next stages of the evolution of graphics mainly occurred in professional industries.



Intel used its 82720 graphics chip as the basis for the $ 1,000 $ Video Graphics Controller Multimode Board (video graphics controller boards with various modes). The device could display data of eight colors with a resolution of 256x256 (or monochrome colors with a resolution of 512x512). 32 kilobytes of display memory was enough to draw lines, arcs, rectangles, and text characters. The chip also had the functions of zoom, split the screen into parts and scroll.



Soon after, SGI released IRIS Graphics for workstations - the GR1.x graphics card with the ability to connect separate (daughter) cards, allowing you to expand the number of colors, use geometry, Z-buffer and overlays / substrates.



At that time, industrial and defense 3D visualization was already quite developed. IBM, General Electric and Martin Marietta (who later bought the GE aerospace department in 1992), together with defense ministry contractors, technology institutes, and NASA, developed various projects that required technology for military and space simulations. Also in 1951, the Navy developed a flight simulator using 3D visualization based on the Whirlwind computer of the Massachusetts Institute of Technology.



In addition to defense contractors, there were companies that supplied professional graphics to the military markets.





Evans & Sutherland, which supplied professional graphics cards like Freedom and REALimage , also supplied graphics for the CT5 flight simulator , which cost $ 20 million and was driven by the DEC PDP-11 mainframe . One of the founders of the company, Ivan Sutherland, developed in 1961 a computer program called Sketchpad, which made it possible to draw geometric shapes using a light pen and display them on a CRT screen in real time.



It was the progenitor of a modern Graphic User Interface (GUI).



In a less closed personal computer industry, a series of 82C43x chips from Chips and Technologies appeared, used in the EGA (Extended Graphics Adapter) and competing with IBM adapters. In 1985, these adapters were installed in many PC / AT clones. This year was also remarkable for the Commodore Amiga computer, released with the OCS chipset. The chipset consisted of three main chips, Agnus, Denise and Paula, which partially reduced the dependence on the CPU in graphics and sound calculations.



In August 1985, three Hong Kong immigrants, Kwok Yuan Ho, Lee Lau, and Benny Lau, established Array Technology Inc. in Canada. By the end of the year, the name was changed to ATI Technologies Inc.



The following year, ATI launched its first product, the OEM Color Emulation Card. It was used to output monochrome green, amber or white phosphor text to the black background of a TTL monitor through a 9-pin DE-9 connector. The card was equipped with 16 KB of memory. She greatly contributed to the fact that in the first year of ATI, the company earned 10 million Canadian dollars. The main reason for this was the contract for the weekly supply of 7,000 chips for Commodore computers.



The advent of color monitors and the lack of common standards led to the formation of the Video Electronics Standards Association (VESA), one of the founding members of which was ATI, as well as NEC and six other manufacturers of graphic adapters.



In 1987, ATI added the Graphics Solution Plus series for OEMs to its product line. It used the IBM PC / XT 8-bit ISA bus for the IBM PC based on the Intel 8086/8088. The chip supported graphical modes MDA, CGA and EGA, configurable using dip-switches. In fact, it was a clone of the Plantronics Colorplus board, but with space for 64 KB of memory. Released in 1987, PEGA1, 1a, and 2a (256KBkB) from Paradise Systems were also Plantronics clones.









ATI EGA 800: 16-color VGA emulation, support for 800x600 resolution



In March, the EGA Wonder series (models 1 to 4) appeared, costing $ 399. It had 256 KB of DRAM and was compatible with CGA, EGA and MDA emulation in resolutions up to 640x350 and with 16 colors. Extended EGA mode was available in series 2,3 and 4.



The high-tech devices were EGA Wonder 800 with 16-color VGA emulation and 800x600 resolution support, and the VGA Improved Performance (VIP) card, which was essentially an EGA Wonder with a DAC module added to provide limited VGA compatibility. The latter cost $ 449 plus $ 99 for the Compaq expansion module.



ATI was far from the only company to ride a wave of growing consumer needs for personal computers.



Many new companies and products appeared that year ... Among them were Trident, SiS, Tamerack, Realtek, Oak Technology, LSI G-2 Inc., Hualon, Cornerstone Imaging and Winbond - they all formed in 1986-87. In the same time period, companies such as AMD, Western Digital / Paradise Systems, Intergraph, Cirrus Logic, Texas Instruments, Gemini, and Genoa began manufacturing their first graphics devices.



Over the next few years, ATI's Wonder Series continued to produce new and exciting products.



In 1988 Small Wonder Graphics Solution was launched with the ability to connect a game controller and composite output (for CGA and MDA emulation), EGA Wonder 480 and 800+ with support for Extended EGA and 16-bit VGA, as well as VGA Wonder and Wonder 16 with Added support for VGA and SVGA.



Wonder 16, equipped with 256 KB of memory, retailed for $ 499, while the 512K version cost $ 699.



In 1989, the improved VGA Wonder / Wonder 16 series appeared, including the VGA Edge 16 (Wonder 1024 series) at a reduced price. New features included a bus-mouse port and support for the VESA Feature Connector. It was a gold-plated connector, similar to a shortened data bus slot connector, cabled to another video controller to bypass an overloaded data bus.



Wonder series updates continued to appear in 1991. The Wonder XL card added VESA 32K and Sierra RAMDAC color compatibility, which increased the maximum display resolution to 640x480 at 72 Hz or 800x600 at 60 Hz. Prices ranged in a wide range: $ 249 for a model with 256 KB, $ 349 for 512 KB, and $ 399 for a version with 1 MB of RAM. The discounted version was called VGA Charger and was based on the Basic-16 of the previous year.









ATI Graphics Ultra ISA (Mach8 + VGA)



ATI created a variant of Wonder XL, on whose extended circuit board is the Creative Sound Blaster 1.5 chip. A device called VGA Stereo-F / X was able to simulate stereo sound from Sound Blaster monaural files roughly as an FM radio.



In May of that year, the Mach series appeared, starting with the Mach8 device. It was sold as a chip or as a board, which allowed using a software interface (AI) to take on part of 2D drawing operations, for example, drawing lines, filling with color, and combining bitmaps (Bit BLIT).



Graphics cards like the ATI VGAWonder GT made it possible to work with 2D and 3D, combining Mach8 with the graphics core (28800-2) of VGA Wonder + cards that performed 3D operations. Thanks to Wonder and Mach8, ATI has surpassed Canadian $ 100 million in revenue, largely due to users switching to Windows 3.0 and the increased workload of 2D operations.



In early 1989, S3 Graphics was created, producing 18 months later its first 2D accelerator chip and S3 911 (or 86C911) graphics card. Important features of the latest are 1 MB VRAM and support for 16-bit color.



In the same year, the S3 911 was replaced by a 924 card - in fact, it was an upgraded version of the 911 with 24-bit color support; the next year it was improved again in version 928, which added support for 32-bit color; accelerators 801 and 805 were also released. The accelerator 801 used the ISA interface, and the 805 used VLB. Between the launch of 911 and the advent of the 3D accelerator, the market was filled with 2D cards for working with GUIs based on the original S3 device. The most notable of these are Tseng labs, Cirrus Logic, Trident, IIT, ATI (Mach32) and Matrox (MAGIC RGB).



In January 1992, Silicon Graphics Inc (SGI) released OpenGL 1.0, a multi-platform, vendor-independent application programming interface (API) for 2D and 3D graphics.



OpenGL has evolved from SGI's own API called IRIS GL (Integrated Raster Imaging System Graphical Library). The idea was to preserve the non-graphical functions of IRIS, and to allow the API to work on systems manufactured by other companies, because competitors appeared on the horizon who supplied their own proprietary APIs.



Initially, OpenGL was aimed at professional UNIX markets, but thanks to the convenience of implementing extensions, it quickly adapted to 3D games.



Microsoft developed its own competing API called Direct3D, and therefore did not make much effort to ensure that OpenGL worked in Windows in the best way.



This situation came to a critical point several years later when id Carmack of id Software, who had previously released Doom, a revolutionary PC game, ported Quake to Windows using OpenGL and openly criticized Direct3D .









Moving to the Future: 1997 GLQuake and Quake Original



Microsoft was adamant - the company refused to license the OpenGL Mini-Client Driver (MCD) for Windows 95, which allowed manufacturers to choose which features can access hardware acceleration. SGI responded to this move by developing an Installable Client Driver (ICD), which not only provided the same functionality, but also did it much better, because MCD only provided rasterization, and ICD added lighting and transform (T&L) .



In the process of developing OpenGL, which initially became popular on workstations, Microsoft closely watched the birth of the gaming market and developed its own proprietary API. In February 1995, the company acquired RenderMorphics, whose Reality Lab API received support from third-party developers and became the core of Direct3D.



At the same time, Brian Hook of 3dfx wrote the Glide API, which would later become the dominant API for games. This was partly due to Microsoft's participation in the Talisman project (tile-based rendering ecosystems), due to which the company had to limit the resources intended for DirectX.



Thanks to the increasing spread of Windows, the D3D interface has become increasingly popular, and proprietary APIs like S3d (S3), Matrox Simple Interface, Creative Graphics Library, C Interface (ATI), SGL (PowerVR), NVLIB (Nvidia), RRedline (Rendition ) and Glide began to lose the love of developers.



They were not even helped by the fact that some of the developers of these proprietary APIs had teamed up with the board manufacturers due to the constant need to expand the list of functions. Among them were an increase in screen resolutions, an increase in color depth (from 16-bit to 24-bit, and then up to 32-bit) and such methods of improving image quality as anti-aliasing. All of these features were needed due to increased throughput, graphics performance, and faster product development cycles.



The year 1993 marked the emergence of new competitors in the graphics market, the most notable of which was Nvidia, founded in January of that year by Jensen Huang, Curtis Reception and Chris Malahowski. Huang was formerly the leader of Coreware at LSI, and Reception and Malahowski came from Sun Microsystems, where they worked on the GX graphics architecture based on SunSPARC .



Soon after, fellow novices Dynamic Pictures, ARK Logic, and Rendition joined Nvidia.



Due to market volatility, many graphic companies have already stopped their business or have been absorbed by competitors. Among them were Tamerack, Gemini Technology, Genoa Systems, Hualon, Headland Technology (purchased by SPEA), Acer, Motorola and Acumos (purchased by Cirrus Logic).



However, ATI remained the only company that went from victory to victory.



In November of the same year, ATI announced the release of the 68890 PC TV decoder chip, which first appeared inside the Video-It card! .. The chip could capture video with a resolution of 320x240 at 15 fps, or 160x120 at 30 fps, as well as compress / decompress it in real time thanks to the integrated Intel i750PD VCP (Video Compression Processor). He could also communicate with the graphics card via the data bus, eliminating the need for additional boards, ports or cables.



Video-It! retailed for $ 399; the product line was also complemented by the Video-Basic model with fewer features.



Five months later, in March, ATI belatedly released a 64-bit accelerator - Mach64.



The fiscal year was not good for ATI, it suffered a loss of CAD 2.7 million, losing market share due to strong competition. Among the competing boards are S3 Vision 968, chosen by many motherboard manufacturers, and Trio64, which received OEM contracts with Dell (Dimension XPS), Compaq (Presario 7170/7180), AT&T (Globalyst), HP (Vectra VE 4) and DEC (Venturis / Celebris).









Vision 968: first S3 video accelerator



Released in 1995, Mach64 was largely the first. It became the first graphic adapter available for both PC and Mac in the Xclaim format (450 or 650 dollars, depending on the amount of memory); like the S3 Trio, it provided accelerated video playback.



Mach64 also became the forerunner of ATI's first professional graphics cards - 3D Pro Turbo and 3D Pro Turbo + PC2TV , which cost $ 599 per version with 2 MB and $ 899 per 4 MB.









ATI Mach64 VT with TV tuner support



The following month, a technology startup called 3DLabs appeared on the scene, born when the DuPont pixel art department bought a division of its parent company along with a GLINT 300SX processor capable of performing OpenGL rendering, fragment processing, and rasterization. Due to the high cost, the cards of the company were initially aimed at the professional market. Fujitsu Sapphire2SX 4MB retails for $ 1,600-2,000, and the 8 megabyte ELSA GLoria 8 for $ 2,600-2,850. However, the 300SX was intended for the gaming market.



In the GLINT 300SX game card, released in 1995, memory was reduced to 2 MB. 1 megabyte was used for textures and Z-buffer, and everything else - for the frame buffer. For 50 extra dollars over the original 349 dollars, the card also had the option of expanding VRAM for compatibility with Direct3D. The product failed to make its way into an already crowded market, but 3DLabs was already working on its descendant in the Permedia series.



At that time, S3 seemed to be everywhere. Trio64 chipsets dominated the high-end OEM market with integrated DACs, a graphics controller, and a clock synthesizer. They used a common frame buffer and supported the hardware functions of video overlays (the allocated part of the graphic memory for rendering the video required by the application). Trio64 and its 32-bit brother Trio32 were sold as OEM devices and individual cards by companies such as Diamond, ELSA, Sparkle, STB, Orchid, Hercules and Number Nine. Diamond Multimedia prices ranged from $ 169 for a ViRGE-based card to $ 569 for Diamond Stealth64 Video with 4 MB VRAM based on Trio64 +.



The mainstream market was also attended by Trident, a long-standing OEM supplier of unpretentious 2D graphics adapters that recently added the 9680 chip to its lineup. The chip boasted most Trio64 features, and the boards usually sold for $ 170-200. In their niche, they provided acceptable 3D performance and good video playback features.



Other newcomers to the mainstream market included Weitek's Power Player 9130 and Alliance Semiconductor's ProMotion 6410 (commonly used with Alaris Matinee or FIS OptiViewPro). Both boards provided excellent scaling at the speed of the CPU, and the latter combined a powerful scaling engine with an anti-blocking scheme for smooth video playback, which was much better than in previous chips like the ATI Mach64, Matrox MGA 2064W and S3 Vision968.









In May, Nvidia released its first NV1 graphics chip, the first commercial GPU capable of 3D rendering, video acceleration, and integrated GUI acceleration.



The company entered into a partnership agreement with ST Microelectronic to manufacture the chip for its 500-nanometer process, and the latter released the STG2000 chip version. Despite the lack of great success, he provided the company with the first profit. Unfortunately for Nvidia, immediately after the start of the first sales by suppliers (in particular, Diamond Edge 3D cards) in September, Microsoft completed work on DirectX 1.0 and released it.



The D3D graphics API was based on rendering triangular polygons, and NV1 used texture mapping on quadrangles. With the driver, limited compatibility with D3D was added: triangles turned into quadrangular surfaces, but the lack of games designed for NV1 doomed it to defeat: the chip was a jack of all trades, but did not achieve superiority in any area.



Most games have been ported with Sega Saturn. NV1 with 4 MB and integrated ports for Saturn (two per expansion card connected to the card by cable) appeared in September 1995 for about $ 450.



Recent changes to Microsoft and the release of the DirectX SDK prevented board manufacturers from gaining direct access to digital video playback. This meant that almost all discrete graphics cards had problems working in Windows 95. The drivers for Win 3.1 from many manufacturers, by contrast, were generally flawless.









In November 1995, ATI announced the release of its first 3D Rage accelerator chip (also known as Mach 64 GT).



The first public demonstration of the chip was held at the E3 video game conference in Los Angeles in May next year. The card itself appeared a month later. 3D Rage combines the Mach64 2D core with 3D features.



Recent edits to the DirectX specs led to 3D Rage having compatibility issues with many games that used this API - mostly no depth buffering. When the frame buffer EDO RAM with 2 MB, the work with 3D was limited to 640x480x16-bit or 400x300x32-bit. Attempting to enable 32-bit color in 600x480 usually damaged the contents of the screen, and the maximum 2D resolution was 1280x1024. In games, the performance was mediocre, but the possibility of full-screen playback of MPEG video to some extent balanced it.



ATI redesigned the chip and Rage II was released in September. He fixed the problems with the D3DX of the first chip and added support for MPEG2 playback. However, the first cards still came with 2 MB of memory, which reduced performance, and also had problems with perspective / geometric transformations. As the device series expanded with Rage II + DVD and 3D Xpression +, the maximum memory capacity increased to 8 MB.



Although ATI was the first to enter the market with a 3D graphics chip, competitors quickly appeared on the scene with other approaches to 3D implementation. In particular, 3dfx, Rendition and VideoLogic.





Screamer 2 released in 1996, running on Windows 95 with the 3dfx Voodoo 1 card



In the race to bring new products to the market, 3Dfx Interactive won Rendition and VideoLogic. However, the performance race was completed before the start - 3Dfx Voodoo Graphics essentially destroyed all competitors.



This is the first of four articles about the history of video processors. In the next part, we recall the glory days of 3Dfx, Rendition, Matrox and a young company called Nvidia.



All Articles