It is evident that this demo is CPU dependent but it scales very
well as a benchmark. With AA settings enabled I was provided with very good
frame rates and excellent visual effects.
Of course, I benchmarked all the maps with no AA/AF just for those of you who do
not use it and for the rest to see maximum frame rates with this card. However,
I was surprised to see that enabling AA did not give a significant hit at 2xAA,
2XQxAA, or 4xAA. Once 6XS and 8xAA was used it was a different story, showing a
hefty hit on frame rates.
UNREAL TOURNAMENT 2004 DEMO ONSLAUGHT
ONS-TORLAN
If this demo is any indication, it appears UT2004, when
released, should play very smooth with any combination of AA and AF as long as
you limit your AA settings to 4x or less. Personally, I can't wait to tryout the
2XQ antialiasing with all quality settings at maximum.
SYNTHETIC BENCHMARKS
A quick run through of some heavily used synthetic benchmarks to wrap up this
review. From the scoring you can compare with results from other reviews and get
a feel how well this card compares. I first used Code Creatures as it stresses
the newer cards sufficiently and requires good bandwidth
Code Creatures
Using complex nature scenes Code Creatures requires DirectX 8 hardware effects
to thoroughly stress even the newest video cards. Although the overall scoring
was consistently in the 4,200 ranges with default settings, I decided to just
show the FPS results at settings of no antialiasing, 2x, and 4x. Tests were
conducted with resolutions of 1024x768, 1280x1024, and 1600x1200.
Code Creatures
At 1024x768, no AA/AF, with default timings of 475/950 frame
rates were consistent at 53.4 FPS. With timings of 550/1000 a six frame per
second increase of 9% was achieved at 58.5 FPS over default timings.
Enabling antialiasing at 2x gives a drop in frame rates at 9% and with 4xAA a
drop of 18% is recorded using default timings.
Overclocking gained 4.2 FPS on the 2xAA test or an 8% increase, and 3.2 FPS in
the 4xAA test or 7% increase. Resolution of 1280x1024 gave an average of 42.5 FPS at 475/950
and 550/1000 gave an extra 4.2 FPS (9%) increase.
With 2xAA frame rates dropped 4.4 FPS (10%) running default timings and with
overclock frame rates dropped 5.6 FPS (12%) from those results recorded with no
antialiasing. Overclocking added 3 FPS (7%) with 2xAA over default timings. With a resolution of 1600/1200 with no AA, 33.7 FPS was recorded
with default timings and overclocking added 3.7 FPS (10%).
Setting AA at 2x produced 29.8 FPS, a 3.9 FPS (12%) hit, while overclocking
increased results 2.9 FPS (9%). At 1600x1200 4xAA dropped frames another 4.1 (14%) from 2xAA while overclocking added 2.4 FPS (9%) over the default 4xAA score.
3DMark2001SE Still one of the most popular DirectX 8 benchmarking tools, FutureMark's
3DMark 2001SE is used by reviewer and gamers alike for comparing performance.
Benchmark testing the XFX FX5950 Ultra was performed in 32-bit mode using pure
hardware DirectX support.
3DMark2001SE Results
3DMark2003
FutureMark's very controversial benchmarking tool with DirectX 9 capability,
3Dmark2003, needs no prior introduction. Seems everyone has an opinion on the
validity of its application. I still maintain that it is an accurate tool in
revealing graphic card performance. However, in this review we limit its use to
the basic test with settings of 1024x768 resolution with only 4xAF applied to
gauge the XFX FX5950 Ultra's performance at default and overclocked timings.
3DMark2003 Results
The XFX FX5950 Ultra performance results are very good as you
would expect from a high end graphics card.
AquaMark3
Using the same Krass engine as found in actual games such as Aquanox 2 and
Spellforce this synthetic benchmark is a step in the right direction for a
gamer's benchmark. It supports such features as Hyper-Threading, Pixel Shader
and Vertex Shader 2.0, and uses a wide range of DirectX 9 effects and features.
Using the basic triscore test the XFX FX5950 Ultra yielded the following
results:
AquaMark3 Results
CONCLUSION
Expense is one thing and performance is another, this card has both. Currently,
selling for just under $400, this card is expensive but if you are seeking top
end performance, as in this case, you get what you pay for. While price is
generally in the top end among the FX5950 Ultra cards, performance-wise, the XFX
FX5950 Ultra is the top end in performance. A potential drawback, other than
price, is that the card is big. It takes up two PCI slots and the sheer size
could be a problem in some mid-size and smaller cases. The sound from the fan is
noticeable under heavy demand but cooling performance is very good, keeping
temperatures in check at maximum load with both default timings of 475/950 and
with overclocked timings of 550/1000.
Card Installed
In every game I played the XFX FX5950 Ultra combined with NVIDIA's ForceWare 55,
version 56.56, demonstrated excellent speed and playability along with superb
graphic capability. Overall, I can say this card has the capability of enhancing
your gaming experience. If the FX5950 Ultra's prices are within your budget then
the XFX FX5950 Ultra is the card to get!
