|
 |
|
HotHardware's Test Setup |
|
Mainstream Monster |
|
How We
Configured Our Test System:
To help fully explain the scores we have listed in the
following benchmarks, we felt it was necessary to first explain
how this system was set up.
We chose Albatron's PX865PE Lite Pro motherboard as the
basis for our tests, as it would cater to the mainstream or
value-minded builder. We started off by manually optimizing
the BIOS settings to the most aggressive system options
available to us. This meant that the memory frequency
was manually set to DDR400 with the CAS timings set to
2-5-2-2. Since the Albatron PX865PE Lite motherboard came with
a form of memory enhancement, we
opted to enable this as well. Lastly, the AGP Aperture
setting was set to 256MB.
Once the BIOS options were set, the hard drive was formatted
as a FAT32 partition, and Windows XP Professional with
Service Pack 1 was installed. After the Windows installation
completed, we installed the latest Intel chipset drivers and
upgraded to DirectX 9.0b. We then installed the
drivers for the rest of the components, using drivers
supplied on Albatron's driver CD, except for the video cards. For the
two Radeon 9600 cards, we installed the
ATi Catalyst drivers that were released simultaneously with
the 9600 XT, which were
version 3.8. For the GeForce FX 5600, we downloaded
and installed Detonator drivers, version 52.13.
Auto-Updating, Hibernation, and System Restore were
disabled, and then we set up a 768MB permanent page file.
We set the visual effects to "best performance" in system
performance to limit any effects these settings would have
on the benchmarks. Graphical settings in the display
control panels were set to the highest level, however, since
we wanted to stress these components the most.
Vertical Sync was always disabled in the OpenGL tabs in the
control panels.
 |
|
Performance Comparisons
With AquaMark3 |
|
DX8
and DX9 Benchmarks |
|
|
 |
Aquamark 3 comes to us by the way of
Massive Development,
and is the first of our tests to be based on a real game
engine. Massive's release of the game Aquanox in
1999 was panned by critics, yet was one of the first to
implement DX8 shaders, which led to the creation of
Aquamark 2, another benchmark used by many reviewers.
Since the Aquamark benchmarks are based on a real gaming
environment, they must support old and new cards alike. Thus, Aquamark 3
utilizes not only DirectX 9 shaders, but DirectX 8 and 7
as well. We ran AQ3 at 1024x768x32 with no
anti-aliasing, then again with 4 and 6 samples, and one
last time with 4 samples but with 8x Anisotropic
Filtering enabled in the drivers as well. As if
that wasn't enough, we really went to stress the video
cards by running the same tests, but this time at
1600x1200. |
In our first full test of the three cards, the Radeons
dominated with the 9600XT overpowering the GeForce FX 5600
by 40% at 1024x768 and 36% at 1600x1200. Applying
anti-aliasing only seems to help widen the gap between ATi
and NVIDIA. Both 9600s lost four frames at 1024x768
when four samples were applied, and seven to eight frames
with 6 samples applied. With the GeForce FX, the
drop-off in frame rate was more pronounced, dropping 5 frames
and 11 frames, respectively. Putting that into a
performance perspective, that's up to a 24% decrease in
frames for each of the 9600s, and more than double that, 57%
for the GeForce FX 5600. What's most noticeable about
the four tests was how the 5600 bounced back when set at
4xAA with 8X Anisotropic Filtering enabled. This
is one area of performance that NVIDIA has obviously
improved from past driver versions.
 |
|
Benchmarks
With Halo |
|
Halo - Finally on the PC, Where it Belongs |
|
|
 |
For many gamers out
there, the release of Halo marks the end of a long wait,
since it was originally released as an Xbox exclusive a
few years back. No additional patches or tweaks
are needed to use Halo as a benchmark, as Gearbox has happily included
all of the information needed in their README files.
The benchmark runs through four of the cut-scenes from
the game, after which the average frame rate is
displayed to the user. We ran the Halo timedemo twice, once at 1024x768 and again at
1280x1024, which are resolutions that we feel most
players will be using for gaming. |
This was our
first look at the Halo benchmark, and it's safe to say that
the 9600 XT easily beat out not only the GeForce FX 5600, but
the 9600 Pro as well. The extra 100MHz boost to the
core speed of the VPU enables the 9600 XT to obtain a 15-20%
increase in performance over the 9600 Pro at both
resolutions. Not bad for a card that is priced
for the mainstream market.
Unreal Tournament 2003 & Splinter Cell Testing
|