Another familiar benchmark utility here from nVidia. NOTE: Keep in mind that the scale on the graphs is 0 ~ 350 FPS, ChamelonMark
results are on a much lower scale so they may appear to be low at first glance
while they are actually on par for a similar system setup.
DUNGEON SIEGE BENCHMARK
Not only does this benchmark add some variety into the real world performance
benchmarking approach, it's also one of my favorite multiplayer games. This
benchmark takes advantage of Dungeon Siege's unique technique of constantly
streaming information across the AGP bus instead of the traditional "wait, load
level, go" and as such is an intensive test on your CPU and video card.
NOTE: The values shown in the graph below are the average, high and low
values are not shown.
In the graph below, the Default Settings indicates that no changes
were made to the game's default video, sound and effects options. Max
Settings indicates that the following settings were changed:
Shadows: All Complex
Texture Filtering: Trilenear
Object Detail: Max
It's difficult to elaborate much on these results in all honesty. This is due
to the fact that a de facto standard hasn't really been established yet. The
"Mine's Faster!" article in the April 2003 issue of PC Gamer is interesting, if
a bit lite. They gloss over the very details regarding the DSBenchmark that I
was hoping to see when they state:
"Please note that we make several custom changes to the files
in this test to ensure that the game runs at full detail with every card on the
market. So, while you can use the downloadable benchmark as-is, your scores may
differ from ours."
Just a bit vague isn't it? Their results grid also isn't clear as to what
tests had AA enabled. Getting into the details wasn't within the scope of that
article though I'm sure, so I may try to contact them directly to get more
information. So, my best guess at exactly what "ensure that the game runs at
full detail" means is in the bulleted list above.
SPLINTER CELL DEMO
Before we get to the obligatory Quake 3 and UT 2003 benchmarks, how about a
nice diversion with no graphs or decimal points? That's doublespeak for,
"frankly, I wasn't satisfied with my own attempts to at devising a benchmarking
approach for this game". For one thing, this is the initial demo that I used and
there's no benchmark utility/procedure that I'm aware of. I tried using FRAPS
during certain paths through the game (that I did the best to duplicate each
time) but it just didn't feel right. So, instead of forcing the issue now I
decided to focus more on my overall impression of performance (extremely vague I
know) as well as giving attention to an interesting graphical anomaly. So, here
This game is a graphical buffet of eye candy. If you haven't seen
it yet, you should. The lighting and shadows are nothing short of amazing. I
have it for my Xbox and I thought that was nice but this sucker really shines on
The overall performance was mixed (keeping in mind that this is the initial
demo, not retail and no patches). For the most part it played well at 1024x768
with the default settings and no AA or AF enabled and FRAPS reported numbers in
the 40's and it was very playable. This is a stealth game after all and quick,
herky-jerky movements (or even running for that matter) are rarely part of the
game. It's even slower paced than Morrowind for the most part if that helps
paint the picture. So, in my opinion, FPS isn't worth getting worked up about in
this game. However, if you do need to run and dodge around quickly (especially
going up/down stairs) then it can get a bit choppy. Maxing out the settings
(mostly dealing with shadows) didn't seem to affect the appearance or performance of the game
too much. Going up to 1600x1200 introduced FPS in the
20's though so that could be considered a bit slow for a game of this pace.
Enabling AA and AF didn't really seem to affect performance much at all in
1024x768 but it does make a big difference in visual quality due to the slower
pace of the game as you take in your surroundings often.
Now for the strange
graphical blip I noticed. In the two 1600x1200 screenshots below, the one on the
top is without AA or AF while the one on the bottom is with 4X AA and 4X AF. See
what appears to be fog in the bottom image? There is no fog in this part of the
game though and you can even detect somewhat of an ordered (albeit fuzzy) grid
which is on top of the entire scene.
NOTE: Clicking on image below will load a
high-quality1600x1200 PNG file. They're over 2MB each so 56kers beware.
Good lighting - No AA - No AF
Bad lighting - 4X AA - 4X AF
I can't claim to know what the cause
of this is, perhaps it is related to an early bug in the game. I ruled out the
Gainward card as I first thought it might be the first (and only) example of a
negative side effect to overclocking this card. However, I dropped the core and
memory settings to factory defaults and still got the same effect. It didn't
appear anywhere else in the demo that I could detect either so it's not a
Below are two more images of a cropped area from the same
images at 200% magnification.