• Choose a category
• All
• Classic-Fiction

For example, take a typical display at 800 × 600 with a fairly low refresh of 70 Hz. The 70 Hz means the display screen is being redrawn 70 times per second. If you use a low color depth of 256 colors, which is 8 bits (28 = 256), you can multiply all of the values together to see how much data per second has to be sent to the display:

800 × 600 × 1 byte × 70 = 33.6 MBps

If you use the same example at 16 million (24-bit) colors, the figure jumps to 100.8 MBps. You might say, “Well, if PCI runs at 132 MBps, it can handle that!” That statement would be true if the PCI bus had nothing else to do but tend to the video card, but almost every system has more than one PCI device, each requiring part of that throughput. The PCI bus simply cannot handle the needs of many current systems.

AGP

Intel answered the desire for video bandwidth even higher than PCI with the Accelerated Graphics Port (AGP). AGP is a single, special port, similar to a PCI slot, that is dedicated to video. You will never see a motherboard with two AGP slots. Figure 19-38 shows an early-generation AGP. AGP is derived from the 66-MHz, 32-bit PCI 2.1 specification. AGP uses a function called strobing that increases the signals two, four, and eight times for each clock cycle.

Figure 19-38 AGP

Simply describing AGP as a faster PCI would seriously misrepresent the power of AGP. AGP has several technological advantages over PCI, including the bus, the internal operations, and the capability to handle 3-D texturing.

First, AGP resides alone, on its own personal data bus, connected directly to the Northbridge (see Figure 19-39). This is important because more advanced versions of AGP outperform every bus on the system except the frontside bus!

Second, AGP takes advantage of pipelining commands, similar to the way CPUs pipeline. Third, AGP has a feature called sidebanding—basically a second data bus that enables the video card to send more commands to the Northbridge while receiving other commands at the same time.

Figure 19-39 An AGP bus

Video cards do all kinds of neat stuff with their RAM; for example, video cards store copies of individual windows so they can display the windows at different points on the screen very quickly. A demanding application can quickly max out the onboard RAM on a video card, so AGP provides a pathway so the AGP card can “steal” chunks of the regular system memory to store video information, especially textures. This is generically called a system memory access and is quite popular.

* * *

NOTE Intel couldn’t quite bring itself to call AGP’s system memory access… err…system memory access, so they use a couple of different terms. The video processor maps out a portion of system memory by using the Graphics Address Remapping Table (GART). The size of the remapped region is called the AGP aperture. A typical AGP aperture is 32 MB or 64 MB.

AGP has gone through three sets of specifications (AGP1.0, AGP2.0, and AGP3.0), but the official names tend to be ignored. Most techs and consumers refer to the various cards by their strobe multiplier, such as AGP 1×, 2×, 4×, and 8×. The only problem with blurring the distinctions between the specifications comes from the fact that many new motherboards simply don’t support the older AGP cards because the older cards require a different physical connection than the new ones.

Some motherboards support multiple types of AGP. Figure 19-40 shows an AGP slot that accommodates everything up to 8×, even the very rare AGP Pro cards. Note that the tab on the slot covers the extra pins required for AGP Pro.

Because many AGP cards will run on older AGP motherboards, you can get away with mixing AGP specifications. To get the best, most stable performance possible, you should use an AGP card that’s fully supported by the motherboard.

Figure 19-40 AGP 8× slot

The only significant downside to AGP lies in the close connection tolerances required by the cards themselves. It’s very common to snap in a new AGP card and power