Home Intro Completeness Contribution Membership Questions Back 
Hardware Level VGA and SVGA Video Programming Information Page
About the FreeVGA Project 

        As a software programmer for the PC architecture I have noticed that there is plenty of free information existing on the Internet for programming PC hardware, with the notable exception of video adapters, MPEG cards and the like. When the VGA was the standard adapter card in PC's, programming was relatively straightforward. However, when SVGA adapters appeared on the market, there was little standardization between separate vendor's products. For this reason, applications and graphical user interface systems required specialized drivers to utilize the extended capabilities of the SVGA adapters. Either these specialized drivers were written by the video card vendor or they had to be written by the application vendor. Unfortunately due to the cost of specifications and complexity of hardware, most free or shareware programs supported only the standard VGA. The goal of this project, under my direction, is to explore the area of low-level video programming and provide information for programmers of free software throughout the world.
        With the growing popularity of the free software concept, more and more specialized applications are written every day. Many of these applications could take advantage of the specialized features of the video hardware but often the information required is not available. This projects goal is to make that information available By gaining the cooperation of both programmers and hardware manufacturers an excellent reference can be developed. Internet technology makes it possible to provide a free resource updated in a timely fashion as opposed to printed matter which takes time to print and deliver.

        Why would a low-level programmer need this reference rather than simply use datasheets and chipset specifications from the manufacturers themselves? First, the manufacturers may no longer provide literature on "obsolete" chipsets. Second, such datasheets are often aimed at hardware designers rather than programmers. Programmers are trying to implement software on existing hardware rather design hardware. Many of the details needed to program the hardware are dependent on the implementation of the video adapter. Datasheets typically only provide information on one particular component of the video hardware. It is necessary to understand how the components in the adapter are "wired" together to be able to program the adapter. This information is rarely found in datasheets or manufacturer's documentation. Third, as demonstrated on the VGA and some specific SVGA hardware, there are always programmers who can find ways to cleverly program the hardware to provide capabilities unimagined by the manufacturer. To do this, it requires the programmer to have intimate knowledge of the hardware, as BIOS services provide a "lowest common denominator" of capability.

        This section is for those interested in contributing to the FreeVGA Project. The primary objective of this project is to gather information about video hardware, to verify this information as best as possible, then to organize the information in a form usable for any application, and finally to make this information freely available to all programmers. Because of the non-profit nature of this project, all information provided is the result of generous contribution by myself or others. The primary resources required by this project are: chipset datasheets/documentation, developer kits for video boards, video adapter boards used for testing and verifying information, and finally "postcards from the bleeding edge" i.e. information about the real world problems and their workarounds from video programmers. If you can provide any of these resources to the project or any other related assistance myself and other programmers who benefit give thanks to your generosity. Your name will be forever enshrined on the list of contributors, along with a link to your homepage if you so desire.
     I can be reached via the feedback form. If you wish to donate hardware or documentation, please send it, along with your name and the link you wish to include in the list of contributors to:

Joshua Neal
FreeVGA Project
925 N. Coronado Dr.
Gilbert, AZ 85234

        Because of the nature of this project, any contributor can consider themselves a member if they wish to do so. As the founder of this project I am willing to donate my time and resources to ensuring the continuing organization, accuracy, and usability of the FreeVGA Project's documentation. I will continue to do this indefinitely, although if the task becomes overwhelming I will solicit volunteers to assist with the project. There may at some point in the future be special considerations for vendors that choose to support the projects goals.

Open-Ended Questions
        The problem of documenting video card operation and behavior is difficult because of the large amount of information available, and because video hardware is constantly evolving, making documentation a problem of hitting a moving target. Another problem is that because video cards are projects of human endeavor and due to their complexity, their implementation often differs from published specifications. Even two manufacturer's products based upon the same chipset can contain enough variation to make them separate cases from a programmer's viewpoint. The FreeVGA Project attempts to provide answers to the following questions:

        Because video hardware was developed by many independent vendors along separate evolutionary paths, there is very little knowledge about how to identify the particular model of video card present in a machine. Because identifying the particular model is crucial to utilizing the advanced features of a specific model, this is an important task that nearly all software written for the SVGA needs to perform. In many cases, such as when writing programs under an operating system other than MS-DOS/Windows, it may not be possible nor is it good practice to access BIOS for determining the specific model. Furthermore, the more recent PCI bus design is being incorporated into many systems with non-80x86 chips, such as the PowerPC, Alpha, and even high end workstations. However, the manufacturers of the hardware may only support Mac and PC versions of their cards, considering other platforms too much of a niche market to support. Since their inception in the PC market, most video cards have had the capability to work with another card (albeit different) in the same machine. Some newer PCI cards allow multiple cards to be used in one machine. Until recently this capability has been unsupported by operating systems and programs. For debugging video routines there is no equal to having a second monitor attached--one monitor can display the program's output while the other provides the debugging interface. Note that the need for a second monitor could be reduced somewhat if better virtualization of the display hardware was implemented. This project aims to give programmers the skills and knowledge to better utilize the video hardware.         BIOS was designed to support MS-DOS programs on the 8086. Computers have progressed far beyond that point but BIOS has remained basically the same (with the exception of VESA BIOS.) Even then, most facilities provided by the video BIOS are not particularly useful. An example is the BIOS Read and Write Pixel commands. These are the only BIOS provided method of accessing video memory other than text functions. Anyone like myself who started learning 8086 assembler to speed up their graphics discovered this function, and said "Cool. This is going to be easy." I then ran my first program and discovered that I had just lost a few orders of magnitude of performance. Thus I started to learn to interact with the video card hardware directly. VESA BIOS is better, although I have seen very few on-board implementations that work properly (if present), usually resulting in the user having to run a TSR program that provides VESA services in RAM. While VESA BIOS does provide some facilities for non-real mode problems it still does place a function call penalty on video code. My biggest complaint is that with VESA BIOS you are restricted to programming to the lowest common denominator of all video hardware, and thus to utilize any special features of a chipset, you still have to learn to program the card directly. Many video cards are now considered "obsolete" by their manufacturer (or the manufacturer has joined the great corporation in the sky...) and developer support is no longer available. The unfortunate problem is that these "obsolete" cards are the same cards being used by non-profit organizations and schools who could otherwise reap the benefits of a wide variety of free software.        This is probably the area where the documentation is most important. These features are programmed differently on each vendor's video hardware, but can improve the performance of a particular application by several orders of magnitude. Programmers such as video game programmers and assembly demo programmers have demonstrated that by pushing hardware to its limits, previously inconceivable animation and special effects are possible. Recent advances in 3D acceleration have made virtual reality on the desktop a possibility. It is crucial to developers that they be able to thoroughly understand the hardware's operation to maximize its performance.       Even programmers programming for the relatively standard VGA hardware can run into problems in this area. This is the reason so many packages state support for IBM VGA or 100% compatible. Frequently programmers encounter slight differences in hardware that, under specific circumstances can cause their programs to fail to operate properly. If programmers had detailed documentation of the hardware differences of specific implementations, programmers could, and are generally willing to, write workarounds in their code in order to provide support for this hardware. Occasionally subtle hardware problems arise in a particular version of a board and is corrected in a later revision (possibly by simply revising the BIOS.) It is important to recognize the earlier version and be able to write software that can deal with its particular problem. In addition, many chipsets are designed in such a way that they can work with a variety of support devices such as clock generators and Video DACs. It is important to know how to detect and control these support devices, which may (and usually is) be different in every implementation. Some of these devices could be interchanged with pin-compatible devices which could provide additional functionality. However, this would require special programming to utilize the device's features.       While testing and verifying the operation of various video boards. I discovered some cards that did not respond properly to my programming. My initial thought was that I was doing something wrong and tried to figure out what was wrong. However, further testing on another identical card demonstrated that the first board had simply failed. There is little diagnostic software for the VGA and SVGA adapters particularly when dealing with some of the more esoteric features. This is primarily because little has been identified about the correct behavior of video cards. Many manufacturers fail to include a thorough diagnostics utility with their hardware, and the diagnostics that are provided are usually specific to one operating system.         This is a particular are of interest to myself and others. This knowledge can be used to create a virtual machine capable of multitasking legacy applications. Particular features that could be provided are the ability to execute a full-screen program in the background, execute a full-screen program in a virtual window on a desktop, emulate a particular video adapter and translate its output to a form compatible with the hardware on the machine, provide the ability to remotely view an applications screen across a network, provide the ability to debug a full-screen application without having to use a dual monitor system or attached text terminal. Huge benefits can be reaped, but all of the details of a particular hardware configuration must be known for proper emulation/virtualization. For example, programs that attempt to autodetect the hardware often rely on undocumented behaviors of video adapters. These undocumented behaviors must be emulated properly for the application to work properly.

Notice: All trademarks used or referred to on this page are the property of their respective owners.
All pages are Copyright © 1997, 1998, J. D. Neal, except where noted. Permission for utilization and distribution is subject to the terms of the FreeVGA Project Copyright License