Performance in Synthetic Benchmarks

Futuremark 3DMark05 build 1.2.0

Not looking very strong in the gaming tests, the Radeon X1900 GT is similar to the Radeon X1800 XT 512MB in 3DMark05 and scores about 1,500 points more than the GeForce 7900 GT although has fewer TMUs and ROPs. It makes sense to analyze the benchmark’s results test by test as they differ a lot among themselves.

The large-scale first test is not a problem for the PowerColor X1900 GT which is on the same level with the Radeon X1800 XT 512MB and far faster than the GeForce 7900 GT. Considering that this test requires a high fill rate from the graphics card, this is an excellent result, although we are not quite sure where it comes from. It probably comes from the architectural peculiarities of the R580 GPU which has more general-purpose registers than the R520 or from some optimizations implemented in ATI’s Catalyst driver.

The second test brings about rather surprising results. The fill rate parameter isn’t at all important here, but the PowerColor X1900 GT somehow cannot compete with the Radeon X1800 XT 512MB as in the previous case, but is considerably slower at the default frequencies. The overclocking helps to nearly eliminate this gap, thus giving us a glimpse of the answer. The performance of the vertex processors seems to determine the overall result of a graphics card in the second test. The Radeon X1800 XT 512MB and the Radeon X1900 GT have the same number of vertex processors, and the GPU frequency growth on the latter card leads to an increase in its vertex processors performance, diminishing the gap between the two products.

The PowerColor X1900 GT (and any other version of the Radeon X1900 GT, too) could do better in the third test, which requires fast processing of complex shaders, if it had more TMUs and ROPs. Anyway, it easily left the Radeon X1800 XL behind and would surely beat the Radeon X1800 XT 512MB if it had 16 texture-mapping units and 16 raster operators. It’s only at the overclocked frequencies that the new card can overtake the latter.

The results of the separate tests agree with the total score, considering the use of full-screen antialiasing. The Radeon X1900 GT is indeed good in 3DMark05, except for the second test where it is slower than the Radeon X1800 XT 512MB.