AGP

The Accelerated Graphics Port (AGP, also called Advanced Graphics Port) is a high-speed point-to-point channel for attaching a single device (generally a graphics card) to a computer's motherboard, primarily to assist in the acceleration of 3D computer graphics. Many classify AGP as a type of computer bus, but this is somewhat of a misnomer since buses generally allow multiple devices to be connected, while AGP does not. Some motherboards have been built with multiple independent AGP slots.

Initial Development

AGP dynamically allocates the PC's normal RAM to store the screen image and to support texture mapping, z-buffering and alpha blending. RAM used in this manner is referred to as the AGP Aperture.

AGP originated from Intel, and it was into a chipset for the Pentium II microprocessor in 1997. AGP cards generally slightly exceed PCI cards in length. AGP became common in mainstream systems in 1998, but PCI video cards are still being produced today.

The first version of AGP, now called AGP 1.0 or AGP 1x, uses a 32-bit channel operating at 66 MHz with 1.5 V or 3.3 V signaling. This results in a maximum data rate for an AGP 1x slot of 266 megabytes per second. In comparison, a standard 32-bit 33 MHz PCI bus (which can be composed of one or more slots) is limited to 133 MB/s.

Newer Versions of AGP

As of 2004, newer versions of AGP increase the transfer rate from two to eight times. Available versions include:

• AGP 2x, using a 32-bit channel operating at 66 MHz double-pumped to an effective 133 MHz resulting in a maximum data rate of 533 megabytes per second; signaling voltages the same as AGP 1x;
• AGP 4x, using a 32-bit channel operating at 133 MHz double-pumped to an effective 266 MHz resulting in a maximum data rate of 1066 megabytes per second; 1.5 V signaling;
• AGP 8x, double-pumped at 266 MHz to give a maximum of 2133 megabytes per second; 0.8 V signaling.

Note: double-pumping means transferring data on both the rising and falling edges of the clock waveform, rather than only on the rising edge as in conventional systems. This technique is also used in double-data-rate SDRAM.

In addition, AGP Pro cards of various types exist. They require more power and are often longer than standard AGP card (though they only connect to one AGP slot). These cards are usually used to accelerate the professional computer-aided design applications employed in the fields of architecture, machining, engineering, and similar fields.

Implementation Details

AGP allows for efficient use of frame buffer memory, thereby improving 2D graphics performance as well. In fact, many RAID systems for "headless" (that is, lacking an attached display) servers plug into the empty AGP slot to take advantage of its increased throughput as opposed to PCI.

AGP provides a coherent memory management design which allows reading scattered data from system memory in rapid bursts. AGP was designed to reduce the overall cost of creating high-end graphics subsystems by using existing system memory. However, general system memory, although cheap, performs much slower than dedicated on-card graphics RAM, and both mid-range and high-end graphics cards rely on their own high-speed RAM for performance. Cheap low-end graphics cards with little on-board RAM benefited from AGP early in the life-cycle of the technology, but the lowered cost of memory since about 2000 has led to even low-end cards having 64MB or 128MB of dedicated RAM, and therefore graphics now rarely utilize system RAM.

End of Life

AGP usage should phase out by around 2005, since Intel has indicated that their future chipsets (scheduled for introduction in mid-2004) will replace AGP support with PCI-Express. NVIDIA's upcoming NV40 will be their last graphics processing unit to natively support AGP, and similarly ATI's upcoming R420 GPU will be their last native AGP chip. Both companies have stated that no further AGP chipsets will be made by them, although they may implement PCI-Express to AGP bridges if necessary to meet market demands.