Fillrate Tests
We'll take a further look at the fillrate capabilities of the X1900 with a few more tests, first looking at the fillrate characteristics with AntiAliaisng enabled.
ATI R580 Measured Fillrate Performance
| X1900 XTX | X1800 XTX | |||||
| Colour Fill | Z Fill | Colour + Z | Colour Fill | Z Fill | Colour + Z | |
| Normal (no AA) | 9719.0 | 10209.8 | 6520.5 | 9447.3 | 9812.2 | 6364.4 |
| 2x | 7849.2 | 9540.4 | 5506.3 | 7650.4 | 9183.0 | 5398.1 |
| 4x | 7069.0 | 4990.4 | 4753.8 | 6885.4 | 4784.0 | 4572.6 |
| 6x | 6417.3 | 3369.7 | 3279.1 | 6253.7 | 3231.3 | 3145.7 |
FSAA fill-rate (Mpixels/sec)
Given that the ROPs of the X1900 are unchanged from X1800 the performance attribute between the two are very similar, and are only differing by their clock rate differences on the core and memory. With 2x AA applied we see that it does have double the fillrate so that its Z and colour fillrates are very similar to the the no AA fillrates, save for a bandwidth induced performance drop. With 4x FSAA the fillrates are halved, as it now takes two cycles to produce a final pixel value, and that is further reduced with 6x FSAA as this mode takes three cycles to produce the 6 samples that will contribute towards the final pixel colour value.
ATI R580 Texture Performance
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | |
| X1900 XTX | 2375.4 | 1906.1 | 1446.3 | 1136.0 | 929.0 | 782.7 | 681.3 | 596.9 |
| X1800 XT | 2339.7 | 1880.0 | 1406.3 | 1104.4 | 894.6 | 753.9 | 651.2 | 572.1 |
| X850 XT PE | 1768.8 | 1355.9 | 1062.9 | 775.5 | 576.9 | 459.3 | 382.8 | 326.8 |
frames per second
ATI R580 Texture Performance Diff 
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | |
| X1900 XTX to X1800 XT | 1.5% | 1.4% | 2.8% | 2.9% | 3.8% | 3.8% | 4.6% | 4.3% |
| X1900 XTX to X850 XT PE | 34.3% | 40.6% | 36.1% | 46.5% | 61.0% | 70.4% | 78.0% | 82.7% |
percentage
The texture performances between the X1900 and X1800 are very similar, with the differences between them close to their clock rate differences, which highlights that the texture sampling capabilities and the scheduler hasn't changed between them. In relation to the X850 the performance at low texture layers is closest to their bandwidth differences, unsurprisingly, but at higher numbers of texture layers the performance difference increases, rather than decreasing towards their texture fillrate differences, by virtue of the better instruction scheduler of the R5xx series and the number of threads it can handle.
Here we'll take a look at the overdraw reduction rates with the GL_Reme application:
ATI R580 Overdraw Performance
| Back to Front | Front to Back | Random | % Improvement from Back to Front | |||
| Front to Back | Random | |||||
| X1900 XTX | Overdraw Factor 3 | 847.1 | 2635.7 | 1389.4 | 211.1% | 64.0% |
| Overdraw Factor 8 | 309.0 | 2260.0 | 899.6 | 631.3% | 191.1% | |
| X1800 XT | Overdraw Factor 3 | 821.9 | 2558.7 | 1355.9 | 211.3% | 65.0% |
| Overdraw Factor 8 | 299.8 | 2190.5 | 871.9 | 630.7% | 190.8% | |
| X850 XT PE | Overdraw Factor 3 | 588.0 | 1854.9 | 935.2 | 215.5% | 59.1% |
| Overdraw Factor 8 | 214.0 | 1628.1 | 630.4 | 660.8% | 194.6% |
frames per second
With the test running at 1600x1200 we see that there is very little overdraw reduction differences between any of the boards here, which is likely due to the fact that all of them already have Hierarchical Z buffers that are optimised for operation at 1600x1200. ATI have stated that they have further increased the size of the Hierarchical Z buffer on X1900 for higher resolution operation than previous boards, however we are presently unable to test such cases.
