14.1. Introduction - 14.2. Characteristic of a graphics card - 14.3. Bus used - 14.4. The graphic circuit - 14.5. Memories - 14.6. The digitizer - analogical - 14.7. 3D possibilities - 14.8. Refresh - 14.9. Video Bios - 14.10. Output connectors - 14.11. Double posting and TV output - 14.12. SLI, Shader, ...
The video is the expander board which transforms a digital data flow into an analogical signal comprehensible by a monitor.
The Hercules video cards were the first boards dedicated to PC computer, monitors were in black and white (or rather according to the amber monitor). The color really appeared with first 286. At that time, the standard was the CGA or EGA, the monitor were to be compatible with the video card. Using a resolution and a number of colors restricted, this standard was replaced by VGA standard wich made it possible to display a resolution of 640 * 480 points 16 colors. A VGA graphics card includes 256 K of memory.
As each manufacturer always wishes to improve, each one left graphics cards S-VGA (for Super VGA). These boards do not meet a standard, each manufacturer develops a driver (a program) who allows the operating system to manage its possibilities. Each driver is specific to the model but also to the operating system. Mode VGA is recognized by all the current operating systems. To pass in higher mode, a driver (the program specific to the board screen) must be installed.
Attention, some Win95 drivers are not compatible with Win98. Nevertheless, as the number of manufacturers of specialized integrated circuits board screen is limited, generally, a driver S3 DX2 (an example of display circuit) is compatible with all the boards including the same graphic chipset, some is the manufacturer.
A zone memory is reserved in the high memory since first 8088. It is currently largely lower than those used by a graphics card (until 256 MB) and can be recovered by orders DOS of high memories. The graphics processor keeps the zones not modified in its own memory. This makes it possible not not to use the processor too much.
A graphic accelerator is characterized by:
Some of these characteristics is identical to that of a microprocessor-based system: processor, memory and Bios. Indeed, under video system is a microprocessor-based system completely in conformity of which the only goal is to process digital data coming from the computer to transfer them towards the converter.
The installation of a graphics card depends inevitably on the bus used (ISA, PCI or for recent boards AGP). The performances depend on it. Nothing is used for to roll in Ferrari on dirt tracks. ISA bus is limited to 16 MB/s, PCI bus to 132 MB/s, AGP 2X with 528 MB/s and the AGP 4X culminates with more 1,7GB/s (in theory).
The graphic circuit (Graphic Process Unit) transforms the digital signals into digital signals organized and readable by the digital converter - analogical. Generally, the circuit is designed to use a certain memory size. For example, S3 TRIO 3d manages a memory of 4 MB, but an extension of 4 additional MB is possible. This is not with advising. Even if the address bus is designed to manage the extension, the interior of the processor is conceived for the basic storage. You authorize better resolutions, but do not improve the total performances of the graphics accelerator.
To speak about the circuits could take many chapters, and considering speed where the video chipset appear and disappear, a weekly update would be necessary, especially on the level 3d. Following information is thus given as an indication.
The principal manufacturers of graphics accelerators are
. Trident (generally into bottom-of-the-range), three-pronged fork 8900 had a long time the appearance of a reference (ISA, 1MB)
. S3. This very active firm generally manufactures into bottom-of-the-range, but can sometimes surprise, like the SAVAGE 4 (basic 16MB, 3d)
. MATROX. The Canadian firm manufactures generally only the top-of-the-range one and their boards are often quoted like reference. Matrox developed in the past its boards screens in 2 versions, a professional version for the image and the technical drawing (old Millenium) and a version plays. Some processors (of oldest with new): G200 - G4000... These boards are seldom the best in 3d, but optimized for the technical drawing, work of images...
. ATI: excellent boards screens of low mediums of range
. INTEL: was interested a little in the graphics boards with the i740. This processor is also integrated in the chipset 1810, managing bases of them 8 MB, it is 3d. This explains why the boards containing I810 do not include a bus AGP. Currently, INTEL is the principal graphic card manufacturer, integrated in the chipset.
. Nvidia: first world manufacturer, all range.
The memory of the graphics card makes it possible to increase the resolution while keeping a number of sizeable colors. For example, a screen of 1024 * 768 pixels 2 colors (black and white) requires 786.432 pixels * 1 bit /8 (for coding in octets)=98 K For 4 colors (Nb Octets = 2), 196,6K and 16 colors (4 bits):384 K This would give for 1600 * 1200 in 16 million colors (14 bits)=30.000 MB, blups...
|Resolution||A number of colors||Memory|
|640 * 480||256||512K|
|32 K||1 MB|
|64 K||1 MB|
|16 Million (24 bits)||1 MB|
|800 * 600||16||512 K|
|32 K||1 MB|
|64 K||1 MB|
|16 M||2 MB|
|1024 * 768||16||512K|
|32 K||2 MB|
|64 K||2 MB|
|16 M||4 MB|
|1280 * 1024||16||1 MB|
|1600 * 1200||256||4 MB|
|64 K||4 MB|
In the case of the boards 3d, a broad part of the memory is used to record textures. A board 3d 32 MB thus does not make it possible to use very high resolutions. In any event, no current screen is able to display the maximum resolutions of the boards screens. For recall, the human eye is limited to 16 Million colors.
The processor is connected to the memory through broad drunk ranging between 64 and 128 bits.
The types of memories used also have their importance.
. the first boards screen used slow memories RAM little expensive.
. Memory VRAM was it little usually used there. The processor and the RAMDAC (digital converter - analogical) can reach it simultaneously.
. Memory WRAM developed by SAMSUNG offers flows 25% superiors to the VRAM and has pre functions decoded such as the drawing of text or the filling of blocks. It appeared with Matrox Millenium.
. The DRAM was developed by MoSYS and consisted of a multitude of small benches reports (32K). This makes it possible to adapt the memory to that necessary per blocks of 32K. For example, a resolution of 1024*768 in true color (24bits) uses 2.5 MB and not 4MB. The flow is here also largely higher than that of the precedents.
. The SGRAM (Synchronous Graphic RAM) has the capacity to function at the same speed that the bus and can reach frequencies higher than 100 MHz. It is to 4 X faster than the DRAM, but is definitely more expensive.
. SDRam allows speeds of 150 MHz
. DDR SDRam, left in 1999 makes it possible to make pass the memory from 150 to 300 MHz, making progress the speed of the boards between 25 and 30 % compared to SDRam. It uses the side going up and going down from the clock signal. Its first use goes back to May 2000 with GeForce 256 of Nvidia 64 MB. This memory is also used in the PC.
The analogical digital converter allows conversion. It is often called RAMDAC. The higher its speed is, the more the vertical frequency of cooling of the screen is high (we will speak again about it). The top-of-the-range graphics cards current ones exceed the 200 MHz.
The current graphics accelerators offer possibilities 3d: the graphics card is able to create directly polygons and fillings (mapping), without direct interventions of the processor of the PC. These possibilities are used mainly for the plays. Without boards 3d (possibly AD one), few possibilities of expressing itself on the new plays. It is on this level that are distinguished the new graphics accelerators compared to those known as 2d.
The operation of these boards require 2 electronic circuits (in more of the memory). With the whole beginning of the boards 3d, the CPU of the PC created the images and an electronic circuit (the board screen) was satisfied as of display them. With the arrival of the first Voodoo chip of 3Dfx, the board included a Triangle Setup Engine and a pixel Rendering engine directly on the graphics card. This made it possible to reduce the use of the CPU of the PC for the creation of polygons and of the effects of pixellisation in the plays. With the arrival of the chip GeForce 256 in 1999, the integrated circuit integrated in more Transformation & Lighting (T&L). The large advantage is that one could work on images made up of more than triangles. For recall, an image 3d entirely consists of triangles and the quality of the image is directly related to the number of triangles. In 2000, GeForce2 can work out until 25 million triangles a second, GeForce 256 managed only 15 million of them.
Put aside technology T&L, the circuits current manage lighting by a shade very elaborate of the pixels: each pixel is equipped individually with data of lighting. GeForce2, for example, can manage 4 pixels simultaneously.
The future graphics accelerators should include anti-alias which makes it possible to smooth the angles, a little as for the printers laser.
This characteristic is related to the RAMDAC. The signal is not sent of a block towards the screen, but points by points according to a line, then that below... The old systems used the mode interlaced to camouflage their weak performances (in opposition to the mode not interlaced Ni). The interlaced principle, used in the TV, refreshes initially only the odd lines, and then takes again the even lines. This causes flutters of the screen, first of all unpleasant, but quickly tiring. A rate of correct cooling begins to 70 Hz. More it is important, the best quality of displaying is. Nevertheless, the old screens do not accept too high frequencies, higher than those minimum of the current boards screens. This explains why it should sometimes be changed screen when graphics card is changed. We will speak again about it.
As well as a PC, a graphics accelerator has its own BIOS. This one is specific to the board manufacturer. This explains differences in performances between 2 identical boards of Chipset. On certain models, the graphic bios is flashable.
The output connector is identical in mode S-VGA (all current boards) to that of standard VGA. Certain boards accept the RGB. In this case, the signal is separate 3 colors (red - green - blue) before being transmitted to the screen via 3 connectors BNC. This requires a special screen. This type of connection is only on the graphic workstations of high range, seldom for the computer.
Now, we found digital connections: DVI and HDMI (witch can transfer also sound).
Displaying on several screens has been used in the workstations dedicated to graphics and technical drawing for several years. Since Win98, you can use 2 boards screen simultaneously (when the drivers allow it) to display either the complete screen on 2 screens CRT, or same images on 2 screens simultaneously (with a board screen PCI and a board screen AGP or screen boards specific). For same simultaneous posting, you can also use specialized apparatuses which one calls of the "splitter", kinds of amplifiers which make it possible to connect until 32 screens simultaneously with the same image.
Certain graphics cards are provided with an exit TV which makes it possible to post your computer screen on a television (on both at the same time or one or the other following the models). This model of graphics accelerator is not really made "to work" on a television set, but well for the DVD, DIVX... The resolution of a television is too weak and the quality of the posting of the data-processing part feels some. On the other hand, the posting of the videos is identical to that of the TV. For recall, the signals TV are taken again in accordance with 3 standards: PAL, SECAM and NTSC. These 3 standards of television are not compatible between them. The emission out of PAL (Belgium, Germany...) towards a TV SECAM (France) and screw poured produces a signal in black and white. System NTSC is used in the United States, Canada... The problem with the graphics cards exit TV is identical. The 3 standards are declined according to various versions which depend on the country.
NViDIA released in 2006 SLI technology. This technology allows to couple two graphics cards of the same type and manufacturer in PCI-Express via a special connector supplied with the same card. The chipset (necessarily a NViDIA NForce) the motherboard must also be compatible with this technology. A single output connector is used, but this function doubles the speed of treatment for compatible games. A Quad version uses 4 connected cards between them, only two are connected in a bus. Some current motherboards accept SLI in 3 channels with 3 ports PCI-express.
Shaders are small pieces of programs run directly by the graphics chipset and not the processor. They are located in DirectX 9.0 and since the GeForce 6600 series by NViDIA. The current version is 3.0. Without this solution, some current games do not work on older PCs. These programs fall into two groups: the Vertex Shader and Pixel Shader. Parts consisting of suites of lines and triangles are treated by the Vertex Shader as the primitive graphics suites. Once these basic graphics decomposed, the graphics chipset will use them to create the color rendering of the different points at the Pixel level.