Specifications
98 PC MAGAZINE NOVEMBER 6, 2007
GeForce 8800 series, nVidia offers the
less-expensive 8600, 8500, and even 8400
cards (although the latter seem to be avail-
able only to OEMs).
These lower-tier cards are priced from
$200 down to as little as $70, but are they
suffi cient for today’s newest technologies?
Can they handle gaming at all?
The Cards
We tested eight cards in total. Three low-
end cards are based on AMD ATI chipsets:
the Radeon HD 2400 Pro, the HD 2600
Pro, and the HD 2600 XT. The other three
are nVidia cards based on the GeForce
8500 GT, the 8600 GT, and the 8600 GTS
chipsets. We tested two high-end cards for
comparison: the Radeon HD 2900 XT and
the GeForce 8800 GTX. These cards cost
signifi cantly more than the other cards in
our roundup, and, as the test results show,
you get what you pay for.
nVidia doesn’t make its own cards: It
produces only the hardware they’re based
Specifi cations are one thing, but the actual performance of
cards can vary. So we tested each of these cards in different
environments to see how they performed. We installed a brand-
new, fresh image of Windows Vista Ultimate on a computer and
turned off variables that might eat up processing time, such as
automatic updates. And we used the latest drivers for each card,
without enabling overclocking.
First we ran 3DMark06, a synthetic test that focuses primar-
ily on the graphics card. It runs six gamelike environments (four
on the graphics card, two on the CPU) and measures the frame
rates that each “game” sustains, as well as other factors, such
as how many polygons and pixels (the building blocks of 3D
images) the card can render each second. 3DMark06 then spits
out a number, in 3DMarks, which constitutes the card’s score.
Not surprisingly, the more a card costs, the higher its score
and the better it generally does. This test shows why hard-core
gamers spend all that money on a high-end GPU: Pixel-pushing
power makes all games look better.
We also ran a number of gaming tests. These are performed
not only for comparison but also to see how well a card can run
a particular 3D game. Most game results are reported in frames
per second (fps), which refers to how many frames of anima-
Graphics Card Performance: How We Tested
High scores are best. 3DMARK06
PREY (fps) COMPANY OF HEROES (fps)
4XAA, 8XAF DX9 DX9 4XAA DX10 DX10 4XAA
Radeon HD 2400 Pro
2,292 2.4 1.9 39.4 27.9 14.6 9.6
Radeon HD 2600 Pro
3,393 36.7 14.2 56.9 43.0 35.7 22.3
Radeon HD 2600 XT
4,953 54.8 21.8 58.6 45.9 39.4 30.6
Radeon HD 2900 XT*
10,321 119.7 84.1 51.7 49.4 42.6 40.4
GeForce 8500 GT
2,215 20.2 14.4 27.1 21.0 14.6 13.1
GeForce 8600 GT
3,969 37.5 22.9 47.8 40.1 30.3 26.5
GeForce 8600 GTS
5,184 50.6 32.1 54.6 53.6 38.2 34.8
GeForce 8800 GTX*
10,707 130.4 93.4 59.2 57.3 54.3 51.7
* High-end cards, tested for comparison. All tests were run at 1,260-by-1,024 resolution. AA—Anti-aliasing; AF—anisotropic fi ltering.
BENCHMARK TESTS
upon. The actual nVidia cards we tested
came from ASUS, EVGA, and XFX—popu-
lar manufacturers that assemble the graph-
ics chips, circuit boards, and memory chips
into products for sale on retail shelves.
But keep in mind that the performance
of the nVidia chips doesn’t vary much
despite different company labels, so any
given GeForce card with similar memory
timings and specifi cations from any com-
pany will show performance similar to that
of the card we tested.
tion the card can draw in a single second. A rate of 30 fps is
considered playable. Anything less and you’ll be able to discern
the changes from one frame to the next.
We ran each test at the relatively low resolution of 1,280-by-
1,024, with the game set at its defaults. We then ran the same
test with some features turned on: namely, anti-aliasing—the
removal of jagged edges and lines—and, if it was available,
anisotropic fi ltering—the smoothing of the image between close
up and faraway items. (Company of Heroes, for example, does
not have anisotropic fi ltering.) These features further burden the
graphics card and therefore result in lower frame rates; however,
with these features turned on, images look better.
The lowest-end cards couldn’t run the DirectX 9 game Prey
at acceptable frame rates. Better cards could—barely. The more-
expensive gaming cards had no trouble with it. We ran Com-
pany of Heroes twice: once in DirectX 9 and once in DirectX 10.
And just look at the difference! DirectX 10 takes a 30 percent or
higher hit on performance. While in DirectX 9, most of the cards
render the game at playable frame rates; in DirectX 10, few do.
Each of these cards has its strengths and weaknesses. As
most of the benchmark test results showed, higher-end cards
produce better performance. If all you’re looking for is simple
graphics and absolutely no gaming fl are,
the cheapest ATI or nVidia cards will
suffi ce. The upside is, these cards are
remarkably quiet and energy effi cient as
well (especially the ATI models).
The more you want gaming or even
Aero performance, the more you’ll have
to spend—and that’s truly the bottom
line. Besides, DirectX 10 isn’t a draw for
the lower end of these cards at all. If
you’re looking forward to future DirectX
10 gaming, you simply have to go high
end. That means buying an expensive
card from nVidia’s GeForce 8 series or
from ATI’s HD 2000 series. Start saving
those pennies!