BFG Technologies Asylum GeForce FX 5900 Ultra Review - Page 4 Of 8
By Clay Angelly - August 4, 2003
BATTLEFIELD 1942 DEMO
BF1942 is a wildly popular game yet there have been relatively few reviews
that use it for real-world benchmarking. Based on my testing I can certainly
understand why this lack of benchmarking is the case.
Rumor has it that BF1942
is very CPU dependent/limited. I cannot confirm or deny this, but I can say that
based on my testing under the following settings the end result was always an
average of ~40 frames per second. I found this to be odd and obviously invalid data.
No AA/No AF
At 1600x1200 with all settings maxed out and using 4X antialiasing (AA) and 8X anisotropic filtering (AF), I found BF1942 to be playable. At lower resolutions and AA / AF settings performance was better, but the difference wasn't dramatic. The only noticable slowdown occurred when I used 8X AA along with 8X AF (still at 1600x1200). At this point it was not what I consider to be playable. Note that these impressions are based on running through a short, pre-determined path where only light AI resistance was encountered on the Tobruk map. Nevertheless, I was impressed. I'd heard grumblings about BF1942 performance on various cards, but I felt that all modes (with exception of 8X AA / 8X AF) were smooth and responsive.
Benchmarks Image Quality Comparisons
With the unreliable benchmarking results, I decided to use BF1942 to concentrate on image quality (IQ) instead. I
used the following settings for all screenshots (from an Axis spawn point on the Tobruk map)
as listed below:
Note: The comparison images are zoomed-in either at 200% or a mind boggling 500%. This was done by taking the original screenshot and zooming in with JASC Paint Shop Pro 8.0 to prevent any distortion or resizing effects/artifacts. I then took a Windows screenshot (Alt+Print Screen), pasted that back into PSP, cropped out the desired selection and then saved the resulting image.
All images are of no AA / AF (left or top) compared to
images of 4X AA and 8X AF (right or bottom)
Highlighted areas of comparison shown below
Notice how the edges are much smoother in the AA
image (on the right) while the interior of the flag maintains the
same sharpness/quality of the original. AF really has little to no
effect in this isolated situation (200% zoom).
This is a great example of how both AA and AF combine to
give you a much higher quality image. Notice how the interior of the vent
in the bottom image is much cleaner/defined than the top thanks to AF (500% zoom).
This is even a better example of how big a difference
AF can make regarding image quality. See how much more defined the
brick is on the wall surface? Also note how there is more detail in
the "grill" area that is cutout in the wall (500% zoom).
Highlighted area of comparison shown below
This is a good overall example of both AA and AF
improvements. AA makes notable differences in the stair rails,
sloping hillside in background and gutter. The AF improvements are
obvious as the foreground sand area is much more defined. However,
the area of sand from the staircase to the wall in background shows
signs of some clipping/banding (500% zoom).
Bottom Line - Battlefield 1942: Since this game is so popular
you'll read opinions on it that cover the entire spectrum. My take on it was
that it ran great other than the menu getting blurry at times.
STAR TREK ELITE FORCE 2 DEMO
Based on a highly modified and enhanced Quake III engine; this game offers a
diversion to the common Quake III benchmarks while still representing the
underlying engine. During my testing I felt that the following settings
available from the game's Advanced Video options menu could cause some confusion
when used together with the NVIDIA control panel settings: "Full Screen
Antialiasing", "nVidia Antialiasing", "Anisotropic Filtering". I decided to
always have these settings unchecked (off) so that there was no doubt that the
game could be overriding the NVIDIA control panel settings.
Full Screen AA
Simple Lighting Distance
Level of Detail
Use Weapon Aim
Unlike BF1942, I was able to get consistent and valid results from my
benchmarking in this game. My scenario for testing was to eliminate all the
aliens, which takes around 2 minutes, from the Attrexian Station hangar (as
Mike did in his 5900 Ultra preview article). I used a balance attack of the
Phaser, Federation Assault Rifle and Enhanced Compression Rifle. I used each
weapons primary and alternate fire modes. This game is gorgeous and I was able
to fully enjoy the eye candy while still having playable frame rates (40 fps) up
to 1024x768 at 8X AA / 8X AF. Once I went higher in resolution at 8X AA the frame rates dropped below acceptable levels.
Bottom Line - Star Trek Elite Force 2: I had a blast with this
game. Some of the higher resolution and AA / AF settings did cause performance hits which I feel are related more to my system than the card
since the early test results were pretty flat across the board.
UNREAL TOURNAMENT 2003 DEMO
There's not much left to be said about UT2K3. The only thing worth mentioning
here is that I used the popular benchmarking utility
UT2KBench 2.0 from
BensCustomCases.com for all test results. The UT2KBench application states
that "This benchmark is always run using maximum detail." I also played through
a variety of resolutions and AA / AF settings with all of the graphical settings
as high as they would go.
I acknowledge all of the attention this game is currently receiving regarding
image quality, pseudo-filtering techniques, etc. However, I didn't test these
claims as it is really out of the scope of this review. My personal opinion is
that the different techniques' end result is tough to distinguish while playing
the game. I don't believe that the ends necessarily justifies the means, though,
but I also think the jury is still out on this one overall.
All of my impressions are based off playing a lot of Antalus and Asbestos bot
matches from anywhere between a few minutes to close to over an hour or so. I
tried a wide variety of resolutions and AA / AF settings and found that anything
from 1024x768 at 8X AA / 8X AF and lower was what I felt to be playable. Anything
higher than 1024x768 at these AA / AF settings dropped the frame rate to below 30
and the hit was just too much. The game looks great at anything higher than
2X AA in my opinion and my favorite setting was 1280x960 at 4X AA / 8X AF.
Benchmarks - UPDATED
Initially, there was a combination of a copy/paste error and some
discrepancies in the frame rate values for the 8XAA / 8XAF settings. I've since
re-run all of the benchmarks with
UT2KBench 2.0 from
BensCustomCases.com and the updated results are reflected below. The frame
rate values for the 8XAA / 8XAF settings actually increased from
my initial benchmarks by about 10-15 fps.
Bottom Line - Unreal Tournament 2003: I came away from this
feeling like I could play it at basically any setting or resolution comfortably.
Despite all the hubbub surrounding this game right now, I really couldn't see
much of anything that bothered me regarding IQ other than some cases where I did
notice that the filtering appeared to be a bit less than perfect. Of course while I was staring at this I was quickly put out of my misery by a bot that was
actually playing the game :)