Evolution of Graphics Display Chips Technology
D.J. Mac Farlane
Professor Springer
Computer Literacy
October 22, 2003
Throughout the last two decades, video display technology has evolved more immensely than one would ever have imagined: its developed exponentially in fact. As one should expect with exponential technological expansion, most of the new developments have occurred within the last six years. In such a short time, we have progressed from monochromes (2 color or black-and-white) capable graphics chips to versatile 3D capable AGP display adapters that produce visually stunning 3D rendered environments in games and CAD engineering based software.
The beginning of Graphics history is very humble. It begins with lowly monochrome display cards that were plugged into ISA (Industrial Standard Architecture) slots produce basically two colors (black & some other color, usually white). Since monochrome cards were decent for rendering text only, the Color Graphics Array standard emerged that could spice up text with up to 4 colors. *16 colors. Later on, the Video Graphics Array capable of rendering a broad pallete of 256 colors truly gave the world what the name implied – the first color pallete sizeable enough to produce full-motion videos which looked were an impressive truly awesome experience at the time. The first massively popular video games like Wolfenstein 3D, DOOM, and Myst were all built on 256 color capable technology.
Unfortunately though, the ISA bus lacked the memory bandwidth necessary to be fast enough to perform all these more complex resource demanding operations that were previously unheard of. Beginning with some of the 486 motherboards, the introduction of VESA Local Bus architecture operations perform better. The VESA Local bus closely resembled what looked like a hybrid PCI & ISA video Adapter, so the slots were very long and required more motherboard space than ISA. The cards themselves were even longer than the massive slot size.
So, not surprisingly this disadvantage was tackled very soon with the introduction of standard PCI ports which were first massively available on Pentium motherboard platforms. Unlike the VESA Local Bus architecture, the PCI architecture was not dedicated to video graphics only in addition to being faster than PCI local bus adapters. So, this massive memory bandwidth increase had a major impact on motherboard platforms industry because the bandwidth could optimize the performance of any device adapter that could plug into it and is still the common medium for most system adapter cards sold on the market today!
With that said, the PCI architecture was arguably the biggest quantum leap ever to be seen in the video display adapter industry as well. PCI cards were much faster than their predecsessors. PCI was an important turning point because not only was the continued development of 2D capable cards, it also witnessed the advent of 3D-only and combo 2D/3D cards. Even since the introduction of AGP cards since 1997, PCI video cards are still developed for the niche market segment of those who have PCI only motherboards. So from here on out, the story of PCI does not diminish.
As briefly referenced before, the first 3D-only cards were made possible by PCI architecture. Many of the 3D-only cards that existed from 1996 to 1998 were much faster than 2D cards and 2D/3D combo cards that existed at this time. 3dfx, founded in 1994, is the one corporation that is responsible for the true turning point in graphics chip design in history. When 3dfx introduced their “Voodoo” 4MB PCI graphics accelerator, the industry would never again be the same. The Voodoo graphics accelerator was capable of rendering polygons at unparalleled speeds. In 3D engine environments, with a little extra programming for the Glide API (Advanced Programming Interface), a derivative of OpenGL, games could run at a fluid 30 frames per second or higher and the textures would be beautifully blended to eliminate the bland and pixilated “squares” that composed each of the textures in games.
The Voodoo Graphics chips isn’t the only remarkable chip technology developed by 3dfx. Because the 3D functionality and rendering features are essentially the same as the original Voodoo, the development of the Voodoo 2 chip was mainly noteworthy more for its exponential gains in performance rather than its outstanding feature set. When launched February, 1998, equipped with a 90 MHZ chip and up to 12MB of EDO memory, a single Voodoo 2 card boasted more than twice the performance of the original Voodoo graphic accelerator. What is unique about the Voodoo 2 based cards is that they are practically the only card ever developed to feature SLI (scan-line interleave) technology combine the graphics power of two separate 3D accelerators in conjunction with one another. To be honest, this SLI was designed by Quantum 3D to use two original Voodoo chips on a single card solution, but were unheard of because they were exorbitantly priced at over $1000 while a single Voodoo could be had for $300 or less! Technically, the Voodoo 2 SLI solution should be considered to be a 3 video card solution, because they required an additional 2D display adapter to render the operating system’s GUI and desktop applications, or the MS-DOS environment for that matter. Another significant trivia fact is that the Voodoo 2 is the 3D only chip that had was on a card that had a design option for the AGP slot. The AGP versions of Voodoo 2 technology were rare and could not be used for SLI because motherboards only have one AGP (Accelerated Graphics Port) port.
As significant as the Voodoo 2 was, it did mark the end of 3D only add-in adapters. However, it didn’t mark the beginning of 2D/3D combo cards either. The development of 2D/3D combo cards coincided with 3D only add-in cards for 2 years. One could even say this development preceded 3D only cards for a few months. The first of these, particularly the S3 Virge and the Rendition Verite chips, were known as 3D “Decelerators” because of their extremely slow 3D rendering performance that were unattractive to gamers. Launched in 1995, the S3 Virge was the first widely available “3D capable” chip on the market. There were only a handful of games that had patches to run on this chip because its 2D performance superceded its 3D rendering performance, so it was actually more preferred to use the 2D side of the chip to render 3D polygon based environments over its visually superior 3D component. The Rendition Verite chip was nearly so bad at 3D performance as the Virge, but, because it was launched very close to the time of the 3dfx Voodoo, this chip never gained a significant market for its 3D features.
In the fall of 1998, a relatively new company to the 3D market, nVidia, unveiled the Riva TNT chip which was a head-to-head competitor against the Voodoo2 series of chips. Although, it did not have superior performance to the Voodoo 2, it managed to outgun the Voodoo2 series in several aspects. The TNT was one of the first 2D/3D capable chips that had design options for the AGP slots to actually be worth its money. Many of the AGP chips of 1997 and 1998 were outperformed by competing PCI products because the AGP architecture was so new developers were not accustomed to it and often developed the products not completely within spec to AGP standards for voltages and design. Since the TNT was almost as fast as the Voodoo 2, the chip became attractive because of its ability to render 32-bit color in 3D engines and its DirectX 6 Design. Yet another attractive point to the TNT is that it was a single card solution while the Voodoo 2 solutions occupied up to 3 slots. The TNT was also the predecessor of the similar TNT 2 chip, which was one of the first chip lines to feature an array of Graphics solutions for virtually every market segment.
Modern 3D chip development begins in 1999 with the introduction of T&L (Transform & Lighting) units to be included on the 3D chips themselves. Ever since nVidia launched its T&L capable GeForce Series they labeled this card and all of their succeeding chips as “GPU” (Graphics Processing Units) because the T&L calculations are now performed directly on the graphics chip as opposed to occupying clock cycles on the CPU itself. Freeing up CPU clock cycles thus enables the CPU to perform more non-graphics calculations inside games such as physics and AI. Even though not directly related to T&L units, ansiotropic filtering and Full Scene Antialiasing which and improves texture clarity long distances and reduces jaggedness of polygons becomes of and coincidentally increasingly demanded by gamers during the DirectX 7 era. In fact the DirectX 7 era continued until 2001 with launched of later products with similar features such as ATi’s Radeon 7000 series and the GeForce2 series of cards.
Vertex Shaders, Pixel Shaders, and programmable T&L based GPUs mark today’s developments graphics cards. Vertex and Pixel shaders have largely been the focus of both DirectX 8, 8.1 & 9 APIs and the latest versions of OpenGL. The essential most basic elements of these developments were first introduced in mid-spring 2001 with the introduction of the GeForce 3 GPU. ATi launched a product with similar features known as the Radeon 8500 series, but was slightly visually superior because its chips was designed to the DirectX 8.1 specification as opposed to the GeForce3 & 4 being designed for DirectX 8.0.
In fall 2002, ATi launched the Radeon 9700 (code named R300 series) card which is by-and-large contains essentially the core developments of even today’s products. Although the R300 series of chips basically are known for improved shaders, the development it is significant because of the massive scale to which Vertex Shaders and Pixel Shaders were redefined. The cards based upon the new Pixel Shader 2.0 and Vertex Shader 2.0 allow for vastly much more engine developers like Gabe Newell at VALVe software and John Carmack at idsoftware much more flexible programming options for game engines than cards based upon prior shader specifications.
So what does the future hold? In this ever expanding technological field, key development occur every year. Next year we should see the introduction of motherboards with the next generation of slot architecture – PCI-X. PCI-X is not to be confused with the older PCI, but like the old PCI is a multipurpose architecture. Also on the horizon is ATi’s R420 series of chips and nVidia’s NV40.
Works Cited
Chezik, Peter. Graphics Cards. <http://216.239.41.104/search?q=cache:G4-du5DatugJ:www.eecis.udel.edu/~humphrey/Graphics Cards.ppt+Graphics+Card+History&hl=en&ie=UTF-8> October 18, 2003.
Donelan, Jenny. Gaming Retorspective. Computer Graphics World. March 2002.
Lal Shimpi, Anand. ATI Radeon 9700 Pro – Delivering as Promised. Anadtech. August 19, 2002. <http://www.anandtech.com/video/showdoc.html?i=1683> October 18, 2003.
Penfold, Don. Vetex Shaders and Pixel Shaders. Tom’s Hardware Guide. January 16, 2002. <http://www.tomshardware.com/graphic/20020116/index.html> October 17, 2003.
I can't post anymore today, I gotta get back to work.
Facebook
Twitter
Instagram