CertiGuide to A+ (A+ 4 Real)  9  Chapter 9: Graphic Cards       9  Card Components, Modes and Architecture

SVGA/XGA 1 Memory

Clearly, the graphics card does not exist within a void. It is connected to the rest of the PC hardware at a deep level. How deep depends on the card.

The PCI bus (an acronym for Peripheral Component Interconnect) was the original interface used to connect graphics cards to the rest of the system, and back in the 1990’s it did the job perfectly. The problem is, with modern applications and games passing huge amounts of data to the graphics card, the PCI bus just is not as fast as it was once considered. In fact, with regard to graphics, the PCI bus is actually quite a bottleneck. So although the PCI bus is more than adequate for basic usage (and in fact is still in wide use), a solution was needed for the ever-growing demand for bandwidth.

In mid-1996, Intel unveiled AGP⁹⁴ (an acronym for Accelerated Graphics Port), releasing it to the market late 1997. Since its release AGP has had many improvements, and has gone from single speed AGP to 8x AGP, which is literally 8 times the speed of the PCI bus - it is actually based on the PCI 2.1 architecture, which was intended to replace the original PCI bus. The underlying technology has remained the same, though. Firstly, the AGP bus has a direct connection to system memory, allowing for fast access by the graphics card for data too large to fit on the onboard memory. This allows for huge amounts of graphic data to be stored and used. One important point is that the frame buffer is always stored on the card's memory, for performance reasons.

Second, unlike PCI, AGP is designed solely for graphics cards and does not share the bandwidth from the system bus with any other devices. Third, whilst the AGP slot is still32 bit, it uses a 64 bit wide data path (as there are contacts on both sides of the port and card). Fourth, AGP uses an extra 88 sideband addresses that allow for simultaneous access and instructions to be transferred. Fifth, AGP transmits data on all the rises and falls of a clock cycle whilst PCI only transmits data once per clock cycle. All of these factors combine to allow 4x AGP to hold data throughput at 1024mb/sec. In theory 8x AGP can provide even better performance, however there is currently little hardware available to take advantage of this power making an 8x AGP card more of an investment in future development than a practical way to improve performance. With motherboards and processors now providing FSB's at over 333 MHz, this potential may soon be realized.

Almost all quality modern motherboards come with an AGP slot, and most operating systems support them. As AGP graphics cards still use the VBE standard, it is possible at least get a VGA mode working with a compatible operating system. AGP cards are designed for their power though, and the correct drivers for the card should always be installed. This is especially important when the card is used for 3D work, as the vendors regularly release updates that improve performance and add extra features.

It should be noted that even though AGP is now the preferred method of connecting graphics cards to a system, PCI cards have not died off totally. With the advent of AGP, PCI has found a new purpose in the graphics world. As motherboards only have one AGP slot, PCI graphics cards are now commonly used to allow dual monitors to be run from one machine. This dual monitor configuration is used a great deal by graphic artists, CAD designers and games developers thanks to the greater flexibility it provides. Not to be outdone, AGP graphics card manufacturers have started producing “dual head” graphics cards, with 2 D-Type monitor connectors (or 1 D-Type and 1 DVI Out) so that a user can take advantage of AGP features on both monitors.

SVGA/XGA 1 Memory