MaximumPC 2005 11

(Dariusz) #1

24 MA XIMUMPC NOVEMBER 2005


BUYERS GUIDE


The acronym
stands for
“graphics pro-
cessing unit,”
which is the
integrated cir-
cuit responsible
for handling 2D
graphics, 3D
rendering, and
even video. Even
the lowest-end
GPUs are suffi cient for tasks such as word processing
and web browsing, but 3D gaming and high-def video
playback demand a great deal more horsepower. A
GPU’s power is determined by the number of pixel
pipelines it’s outfi tted with, the width of its interface to
memory, its clock speeds, and support for advanced
shader models.
ATI and nVidia own the market for game-oriented
videocards with their Radeon X800 and GeForce
6- and 7-series cards, respectively. nVidia sells its
GPUs only to third-party vendors, who manufacture
and market videocards to OEMs and consumers.
ATI offers retail boards in addition to selling its GPUs
to OEMs and third-party vendors. Both companies
provide ongoing support to end users by offering free
driver upgrades that fi x bugs and squeeze additional
performance from the hardware. There are several fac-
tors you need to consider when you’re choosing the
GPU to buy.

GPU


Microsoft developed
this collection of APIs
to simplify game devel-
opment and eliminate
compatibility issues
with the company’s
Windows operating
system. Support for
DirectX is ubiquitous in
games, but make sure
the card you’re con-
sidering supports the
latest version (DirectX
9.0c as of this writing).

DIRECTX


CLOCK SPEEDS


When you’re shopping for a new videocard, it pays to be well informed. The
industry has a penchant for fl inging gobs of hype and nonsense, but we can
show you which videocard specs and features really matter, and which are just
marketing baloney.
Finding the right card for your needs and budget requires a fi rm understand-
ing of these specs and features. You gotta know where to make smart trade-
offs. If your budget is tight, for example, you might be better served by
a videocard with slower memory and core clock speeds but more pixel
pipelines than by a videocard with fast clock speeds and fewer pipes.
Here are the most important features to look for.

There are two important clock
speeds on your videocard. There’s
one clock for the GPU (sometimes
referred to as the “core clock”) and
a second clock for the onboard
memory. Clock speeds are mea-
sured in millions of cycles per sec-
ond and are stated in megaHertz
(MHz). Within a GPU family, and with
all else being equal (especially the
number of pixel pipelines), the card
with the faster clock speed will usu-
ally be faster than the card with the
slower clock. Comparing the clock
speeds of different GPU families,
however—ATI’s Radeon series and

nVidia’s GeForce series, for exam-
ple—is not necessarily a reliable
indicator of videocard performance,
especially if the two GPUs have a
different number of pipelines.
If you encounter a videocard with
what seems like an outrageously
high memory clock speed—1GHz or
higher—the manufacturer is probably
publishing the memory’s “effective”
clock speed. DDR RAM is capable of
transmitting data on both the rising
and falling edges of the clock speed,
so DDR RAM clocked at 500MHz
has an effective clock speed of
1000MHz, or 1GHz.

Your GPU executes a series of instruc-
tions to apply textures and transfor-
mations (transparency, deformation,
refl ections, and so on) to the pixels
that make up the onscreen image.
This series of instructions is called a
pipeline, and a stream of pixels is con-
stantly pushed through it. The upshot
is that the GPU handles millions of pix-
els every second. Modern GPUs have
multiple pipelines operating in parallel,
so generally speaking, the more pipes
the GPU has, the faster it will be able
to render an image.

PIXEL PIPELINES


This is a composite label for two components of Microsoft’s
DirectX technology: Pixel Shader Model 3.0 and Vertex Shader
Model 3.0. Shader Model 3.
defi nes a specifi c set of func-
tions and features that your
GPU’s programmable shader
units support. Small shader
programs run in the shader
units, and can defi ne the
surface properties of objects
in games. A more sophisticated shader
model will let game developers cre-
ate complex and realistic environments
while simultaneously reducing CPU and
memory overhead.
nVidia’s GeForce 6- and 7-series
GPUs support Shader Model 3.0 now.
ATI’s upcoming GPU—code-named
R520—is expected to support the tech-
nology when it comes to market later
this year. Shader Model 3.0 hardware
supports higher precision and more
advanced conditional operations than
earlier revisions. Some companies claim
that a card must support Shader Model
3.0 in order to deliver high dynamic-
range lighting in games, but that’s not
actually the case. Support for Shader
Model 3.0, therefore, is much less impor-
tant than most other features.

SHADER MODEL 3.


VIDEOCARD SHOWDOWN

Free download pdf