(^260) PC Hardware: A Beginner’s Guide
Video Processor
On most older video cards, the PC’s CPU is also the video processor and performs all of
the geometric and mathematical calculations of the transform and lighting phases. The
CPU sends the raw screen image to the video card’s frame buffer (the video card’s mem-
ory), from which the video card reads it, performs the setup phase, and writes it back for
the RAMDAC to use.
On newer video cards, the transform and lighting phases are performed on the video
card by its processor, which is also called the graphics processing unit, or GPU. The CPU
extracts the graphics instructions from the application software’s data stream and passes it to
theGPUovertheinterfacebusinuse(eitherAGPorPCI).TheGPUperformsthecalculations
requiredtoproducethedataneededforthesetupphase.LikethedataprocessedontheCPU,
this data is written to the video card’s memory for use in the setu p phase. Regardless of
which processor performs the transform and lighting phases (the CPU or the GPU), there is
much more information produced in these calculations than is received from the application.
When the GPU performs this task, there is less data transmitted over the system bus, which
further reduces the workload of the CPU. Because it has no other responsibilities, the GPU is
able to process the graphics information about 10 times faster than the CPU.
Video Memory
A certain amount of memory is required to hold the graphics information being passed to
the setup phase from the transform and lighting phases. The amount of memory needed
is directly related to the amount of graphics information being passed, the resolution of
the monitor, and the number of graphic dimensions being generated. For example, a
monochrome text display on an MDA monitor required less than 2KB of space, but today’s
3D high-resolution displays can use as much as 64MB of video RAM.
Like the video processor, the location of the memory used to store the graphics informa-
tion has changed as well. The 2KB of memory used by an MDA display was carved out of the
UpperMemoryAreainthePC’sRAM.ThiswasappropriateatthetimebecausetheCPUdid
most of the processing for monochrome text graphics. Working out of system memory was
convenient and, at the time, less expensive than putting RAM on the video card. However
because the need for video memory increased from kilobytes to megabytes and there was a
need for faster data transfers, video memory, more commonly called video RAM, is now
located on the video card along with the GPU, which performs most of the processing.
In some less-expensive home PCs, some of the video processing functions are inte-
grated into the motherboard, and the frame buffer is located in the system RAM. This ap-
proachtovideoRAMiscalledunifiedmemoryarchitecture,whichmeansthatthesystem
RAM is being used to support video along with everything else running on the PC. This
design eliminates the need for a separate video card, along with its cost. This approach
produces a lower quality video compared to those supported directly by a video card
with its own video RAM and processor. The other problem with this design approach is
that if the video system fails, the entire motherboard must be replaced.