Get the latest reviews and tutorials!
Register for the free
Hardware Update Newsletter!




 - Most Active Threads
 - Technical Support
 - CPUs & Overclocking


RSS feed

Sitemap








Video cards

Gigabyte Radeon HD 3870 512MB Review - Page 7

By Vince Freeman May 14, 2008

Unreal Tournament 3 Performance

Unreal Tournament 3 is the newest in a long series of UT-based games, and uses the 3rd-generation Unreal engine. This makes it a definite benchmark contender, as its next-gen graphics engine can really push even high-end cards to the limit, especially at higher detail levels. The only problem is that it includes the standard Botmatch and Flyby in-game tests, neither of which truly replicates the in-game experience from a graphics card perspective.

Botmatch is a bit CPU-oriented and offers more varied scores, and while Flyby does isolate the video component and provides more consistent results, it is not indicative of real-world gaming. So how do we solve the problem? By testing with both, of course.

Unreal Tournament 3 Botmatch Performance

The first benchmark is of a Botmatch run in the WAR-MarketDistrict map, which is chosen due to its heavy in-game battles and sprawling indoor and outdoor locations. Because of some variance in UT3 Botmatch scores, we ran the test 4 times, eliminated the low mark, and then averaged the rest. This provides a reliable score, and is repeatable through subsequent runs. The test resolutions remain at 1280x1024 and 1600x1200, while game details have been set at 2.

The Gigabyte Radeon HD 3870 512MB does quite well at our first UT 3 benchmark test, and although it doesn't pace the field, we're still looking at performance that is competitive with the Radeon HD 2900 XT and GeForce 8800 GT.

Unreal Tournament 3 Botmatch "High Detail" Performance

This next test scenario remains much the same as in other benchmark tests, but as we found some issues with driver anti-aliasing and anisotropic filtering results, we have instead ramped up the UT3 detail settings to 5 (max) and used the standard 1680x1050 and 1920x1200 widescreen resolutions.

The shift to maximum details drops the Gigabyte card back into the pack, pushing it more in line with the GeForce 8800 GTS 320MB.

Unreal Tournament 3 Flyby Performance

The UT3 Flyby benchmarking is the same format as the Botmatch testing, and is done on the WAR-MarketDistrict map, at 1280x1024 and 1600x1200, with "2" detail settings. Due to its test format, the Flyby scores are higher, and more centered on pure graphic performance, but do not include real in-game battle sequences. It's a different way to look at UT3 performance, where relative position may be more important than pure framerates.

The first round of UT3 Flyby benchmarking does not show the Gigabyte Radeon HD 3870 512MB in a good light, and it's not that far ahead of the Radeon HD 3850. It seems that the lower detail settings are not pushing the HD 3870 architecture, and we'll have to see if the higher-resolution/detail testing changes the situation.

Unreal Tournament 3 Flyby "High Detail" Performance

It's the same basic format here as in the Botmatch testing, and UT3 Flyby "High Detail" benchmarking increases the detail settings up to 5 (max) and uses the more demanding 1680x1050 and 1920x1200 widescreen resolutions.

There is some improvement here, but the Gigabyte board still finds itself in the middle of the UT 3 pack. At 1920x1200, it comes extremely close to matching the GeForce 8800 GTS 320MB, but even the Radeon HD 2900 XT shows better legs.


Page 1 Gigabyte Radeon HD 3870 512MB Review
Page 2 Test Setup and Benchmark Software
Page 3 Farcry and Quake 4 Performance
Page 4 F.E.A.R. and Company of Heroes Performance
Page 5 Lost Planet and Supreme Commander Performance
Page 6 Crysis and World in Conflict Performance
  • Page 7 Unreal Tournament 3 Botmatch and Flyby Performance
    Page 8 3DMark06 Advanced Feature Performance
    Page 9 Benchmark Analysis, Overclocking, Value, and Conclusion

    Comment and Contribute



      (Maximum characters: 1200). You have 1200 characters left.