AGP
Accelerated Graphics Port. An Intel-designed port dedicated to graphics. Available on
Pentium II systems only, in Slot 1 motherboards, AGP moves the graphics card off the PCI
bus, and gives it a direct connect to system memory.
PCI
Peripheral Connect Interface. The name given to the expansion bus slots found in the back
of almost all PCs. Hard disk controllers, network cards, audio boards, and graphics cards
are all the different peripherals using the PCI bus. Information from the peripheral to
and from the system memory has to go over the PCI bus, and all these devices individually
contend for bandwidth on the bus.
AGP 1X
Most of today's AGP systems support AGP 1X transfer rates. The rate of information
transfer is twice as fast as PCI transfer rates (256 Mbytes/s on AGP 1X versus 133
Mbytes/s peak on the PCI bus).
AGP 2X
Most systems built in 1998 will support the new AGP 2X standard, which will double the
bandwidth of existing AGP solutions to 512 Mbytes/s.
Direct Memory Access (DMA)
In DMA bitmaps or textures are stored in system or AGP memory and transferred to local,
on-board memory before being processed.
Direct Memory Execute (DME)
In DME, texture information is retrieved from system memory and processed directly to the
screen, decreasing total processing time versus DMA.
Beyond AGP 2X (what is to be expected in the future)
In 1999, Intel expects to be able to support AGP 4X, and achieve transfer rates of 1
Gigabyte/s bandwidth performance.
Sideband signalling
An extra 8-bits of addressing capability built into AGP which, in effect, allows the AGP
graphics board to request information over AGP at the same time as it is receiving data
over the 32-bit datapath of the bus. This is yet another way that AGP graphics board can
create better efficiencies and improve overall graphics performance.
Pipelining
AGP graphics boards can queue multiple commands using a method called pipelining. Without
pipelining, commands must be received one at a time.
Dual Independent Bus (DIB)
Pentium II AGP system CPUs have the advantage of keeping the paths to local cache and
system memory separate. Because of the DIB architecture the CPU gets a performance boost
by being able to access its local cache independently of how it is working with system
memory. The result is that while the CPU is working on processing graphics information, it
is passing on the results of previous calculations to system memory to be transferred
across the Accelerated Graphics Port directly to the graphics board.
Rasterization
The act of rendering 3D images to a 2D screen is called rasterization. Real-time 3D
graphics boards take 3D coordinates from the CPU for a given frame of animation and render
them to the screen, calculating perspective, viewpoint, and visibility of objects as well
as rendering surfaces and textures on to surfaces.
Scalability
An often used term in computing that simply means improvements in software performance
should mirror improvements in system power. Therefore, as software developers learn more
about the ways in which to use pipelining, DIB, and AGP techniques they will be able to
increase the performance of their applications to scale with the performance of the
hardware.