|
I'm sure you all know the drill here. We'll start
off with 3DMark 2001 SE v330, antialiasing off.

After 640x480, the 5200 could no longer compete. How about we
turn on 4-sample antialiasing?

Again, the 5200 holds up initially, but begins to lose ground
after 640x480. The performance gap is much wider with antialiasing
on, a trend that will continue throughout these benchmarks.
It's been my experience that the Comanche 4 benchmarking tool
is very CPU-bound, but we can still see the tremendous difference
here, especially with 4x AA on.
Close to a 20 FPS difference at 1600x1200? Ouch. I ran that one
a few times just to double-check, and sure enough, the 9000 trounced
the FX every time.
This is the first time I've used Codecreatures in a video card
review and I must say it is a very demanding benchmark. A single
score is generated by averaging the score of three different resolutions;
1024x768, 1280x1024, and 1600x1200.

A fairly close race, but the 9000 inched its way past.
GLExcess is a great tool for measuringL performance, a
very important video card quality for us Half-Life and Quake players.

Other than 1024x768, the 9000 came out on top.
Since 3DMark 2003 is meant to primarily test DX9 cards, and the
Radeon 9000 does not support the DX9 features that the 5200 does,
this benchmark was only used to demonstrate the heavy performance
drop experienced by the 5200 when antialiasing is activated.
For anyone looking to use this card for non-gaming 3D applications,
I've included SPEC Viewperf 7.1. It's another photo finish, with
the 9000 squeaking by with better performance.

We'll round-out the benchmarks with the Unreal Performance Test.
|