LOSTCIRCUITS

Unreal Tournament 2003 Demo

We used the built-in Benchmark at 1280 x 960 at both the balanced and the Highest quality setting:

Balanced

"Balanced" setting: The chart nicely shows that in the botmatch (red columns) being mostly processor limited, overclocking the graphics card makes no difference. For the cumulative Flyby score, the difference is very obvious, though (blue columns).

Quality

When running at the Quality setting, we ran into one issue, that is, the benchmark returned two entirely different sets of result. Even in consecutive runs, we found that the results showed a strictly bimodal distribution, most likely being caused by the drivers not accepting one or the other setting of the quality specifications, that is, changing the mode of AA or AF on the fly. We observed the exact same behavior with both the ATI RADEON and the TYAN TACHYON and since we are only comparing the two cards internally, it does not matter, we simply have two sets of data that both show very similar behavioral patterns.

Average fps for Flyby (blue) and Botmatch (red) (lower score group, which we believe is the correct one). The numbers for the RADEON and TACHYON at identical frequencies were the same, the higher frequency settings are naturally the scores of the TACHYON since it was not possible to run the RADEON that high.

Same as above but the results within the higher score group are shown. R: RADEON, T: TACHYON.

Regardless of the absolute scores, both groups show very good scaling of performance with increasing core and memory frequency.

Next Page:    => Conclusion =>