Creative 3D Blaster 5 FX5900 Ultra Review - Page 5 Of 7
By Steve Angelly & Mike Chambers - September 11, 2003
UNREAL TOURNAMENT 2003
I've included a number of benchmark results from Unreal Tournament 2003, which are based on the Flyby test using the following settings:
(3) Resolutions - 1024x768, 1280x1204 and 1600x1200
(6) Settings - No AA/No AF, 2X AA/8X AF, 4X AA/8X AF, 4XS AA/8X, 6X AA/8X and 8X AA/8X AF
(7) Maps - Antalus, Asbestos, Citadel, Face3, Inferno, Phobos2 and Sun Temple
Results at each resolution are followed with a table showing the percent impact antialiasing had on performance. The final series of test results show the increased performance that was achieved by overclocking.
At a resolution of 1024x768, Flyby performance on the 3D Blaster 5 FX5900 Ultra with No AA/No AF ranges from 157 fps on the Citadel map to 221 fps on Asbestos. The difference in performance when playing UT2003 on an outdoor map compared to an indoor map can be large.
Generally speaking, you should fine-tune the graphics settings in UT2003, and other games, to support the worst case scenario (i.e., an outdoor map with a large number of players).
UT2003 Flyby Benchmark Results - 1024x768
The transition from No AA/No AF to 2X AA/8X AF results in an average loss of 35% in five of the seven maps with Face3 (14%) and Citadel (12%) being the least impacted. However, note that the loss in performance is less than 5% in each map when increasing the level of antialiasing from 2X to 4X. Moving from 4X AA to 4XS AA and higher results in another round of massive hits.
The following table illustrates the percent impact on performance when increasing the level of antialiasing from the previous level.
UT2003 Flyby Antialiasing Impact - 1024x768
For example, moving from 2X to 4X antialiasing on the Inferno map caused a 7% drop in the average frame rate. Moving from 4X to 4XS antialiasing on the same map caused at 46% drop.
UT2003 - Citadel
1280x1024 Resolution
It's been my experience that actual gameplay performance can be estimated by dividing the Fly results by two. Although these results are from 1280x1024, using 4X AA/8X AF at 1280x960 is certainly a possibility with the 3D Blaster 5 FX5900 Ultra.
UT2003 Flyby Benchmark Results - 1280x1024
As expected, the relative impact of antialiasing on performance increases across the board as the resolution increases. At the same time, the amount of memory required by enabling antialiasing at higher resolutions also increases. For example, using 4X AA at a resolution of 1024x768 requires approximately 38 MB of memory while moving to 1280x1024 increases the requirement to 83 MB.
UT2003 Flyby Antialiasing Impact - 1280x1024
Today's high-end graphics cards are extremely powerful as they are able to render a frame of data in 1/60th of a second with AA/AF enabled.
UT2003 - Phobos2
1600x1200 Resolution
It wasn't too long ago that gamers rejoiced when using 4X AA/8X AF at a resolution of 1024x768 became a reality. But we'll have to wait a few hardware generations before 4X AA becomes commonplace at 1600x1200.
UT2003 Flyby Benchmark Results - 1600x1200
UT2003 Flyby Antialiasing Impact - 1600x1200
1600x1200 Overclocking
Then again, we might be able to shave off some time by accelerating the rate at which core and memory frequencies are increasing. For this test, I overclocked the 3D Blaster FX5900 Ultra to 515MHz/950MHz and performed the 1600x1200 benchmarks again.
UT2003 Flyby Overclocking Results - 1600x1200
Even with the 10-15% increase in performance from overclocking, were still a good 30% lower than the results with 4X AA at 1280x1024.
UT2003 Flyby Overclocking Impact - 1600x1200
The 3D Blaster 5 FX5900 Ultra offers an excellent gaming experience in UT2003. Frame rates, with use of AA and AF at Quality settings, are at a level that produces a smooth feel to this fast paced game under a variety of graphics settings. Those of you with 19 or 21-inch monitors will be appreciative of being able to crank up the resolution as I found 1600x1200 with 2X AA/8X AF a setting that produced good performance for me throughout UT2003.