By Clay Angelly - October 23, 2003

UNREAL TOURNAMENT 2003

The results below are from two separate custom time demos created by myself (DM-TokaraForest) and Tim "The Baron" (DM-Icetomb) here at nV News. The DM-TokaraForest time demo isn't nearly as intensive as Tim's DM-Icetomb. Tim summarizes his demo best by saying that:

Mine is more intensive than any custom demo I've ever seen... seems like it needs at least a 1.8Ghz box to be not CPU-dependent, though. But, after that point, it's pure GPU-killing fun.

My DM-TokaraForest consisted of 12 bots, 12 kill limit with a running time of approximately 2 minutes. Tim's DM-Icetomb was a 16 kill limit with a running time of approximately 1 minute 20 seconds (time approximations based on 1024x768 NoAA, NoAF).

Update 10.24.03
Initially I did not include the benchmarks for the BFG 5700 Ultra's default clock speeds (non-overclocked, factory default settings which are 475/900). This graph below shows the results in Unreal Tournament 2003 at those default speeds. Please note that these results are just of the BFG 5700 Ultra since the ATI 9800Pro remained at default clock speeds throughout this review.

Unreal Tournament 2003 (BFG 5700 Ultra @ Default Clock Speeds)

Now that we've established the results at the default clock speeds let's move on to the same benchmarks only now with the BFG 5700 Ultra overclocked to 545/950. The ATI 9800Pro (at its default clock speeds) is also in the mix now.

Unreal Tournament 2003 (BFG 5700 Ultra Overclocked to 545/950)

What's obviously nice about using custom time demos is that you remove the possibility for any special enhancements in the drivers by ATI or NVIDIA. These two offer a good coverage of scenarios that you might encounter yourself regarding CPU and/or GPU limiting factors.

Update: NVIDIA drivers still do not allow for full tri-linear filtering. As a result, an asterisk noting this fact should be kept in mind regarding the benchmark results. There's really no way of knowing what effect this has on the resulting frames per second since full tri-linear filtering cannot be accomplished in Unreal Tournament 2003 with NVIDIA drivers and therefore no comparison can be drawn. Most "best guesses" that I'm aware of estimate this to be anywhere from a few frames per second to ten or more depending on the map, system, etc.

Of special interest to me is that the BFG 5700 Ultra actually beats the ATI Radeon 9800Pro by 10% in DM-TokaraForest at 1024x768 with NoAA/NoAF.

The primary conclusion to be drawn from these results, however, is that the average percentage hit that the BFG 5700 Ultra suffers to the ATI Radeon 9800Pro is approximately 35%. Keeping that in mind, and knowing that BFG 5700 Ultra will be available at retail outlets for $200, the lowest priced ATI Radeon 9800Pro from NewEgg as of 10.23.2003 is $365 (before shipping). That price difference comes out to the ATI Radeon 9800Pro being 45% more expensive than the BFG 5700 Ultra.

Everyone places value differently and admittedly this price to performance ratio I'm presenting is somewhat disputable from various perspectives. At the end of the day though, a BFG 5700 Ultra owner would find themselves in a satisfying situation with very playable frame rates up to 1280x960 and with an extra $165 in their pocket compared to the enthusiast market product offerings.

X²-THE THREAT

X²-The Threat

This gorgeous, upcoming game makes tremendous use of bump mapping, high-polygon counts and DirectX 9. It stresses even the most powerful cards available today. The following settings were used for each test:

Anti-Aliasing: On
Bumpmapping: On
Realtime Shadows: On

NOTE: AA and AF settings were set to application-controlled via each card's driver control panel so that the game's benchmark settings were respected.