Benchmarks - Theoretical Performance
Before going on to look at any actual benchmark scores we'll take a look at the theoretical rates of the boards we're using in this test alongside the X700 XT and PRO:
Core Clock (MHz) | Fill-rate (Mp/s) | Texture Fill-rate (Mt/s) | Triangle (Mtris/p) | Memory Clock (MHz) | Memory Bandwidth (GB/s) | |
X700 XT | 475 | 3800 | 3800 | 713 | 525 | 16.8 |
X700 PRO | 425 | 3400 | 3400 | 638 | 432 | 13.8 |
X800 XT | 500 | 8000 | 8000 | 750 | 500 | 32.0 |
X600 XT | 500 | 2000 | 2000 | 250 | 370 | 11.8 |
9800 PRO | 380 | 3040 | 3040 | 380 | 340 | 21.8 |
X700 XT % Faster Than: | Core Clock | Fill-rate | Texture Fill-rate | Triangle Rate | Memory Clock | Memory Bandwidth |
X700 PRO | 11.8% | 11.8% | 11.8% | 11.8% | 21.5% | 21.5% |
X800 XT | -5.0% | -52.5% | -52.5% | -5.0% | 5.0% | -47.5% |
X600 XT | -5.0% | 90.0% | 90.0% | 185.0% | 41.9% | 41.9% |
9800 PRO | 25.0% | 25.0% | 25.0% | 87.5% | 54.4% | -22.8% |
X700 PRO % Faster Than: | Core Clock | Fill-rate | Texture Fill-rate | Triangle Rate | Memory Clock | Memory Bandwidth |
X800 XT | -15.0% | -57.5% | -57.5% | -15.0% | -16.9% | -58.5% |
X600 XT | -15.0% | 70.0% | 70.0% | 155.0% | 16.8% | 16.8% |
9800 PRO | 11.8% | 11.8% | 11.8% | 67.8% | 27.1% | -36.5% |
Comparing the XT and the PRO, as they both utilise the same core, the theoretical rates purely fall in-line with their differing clockspeeds and as such the XT has a 12% fill-rate and geometry rate andvantage and 22% greater memory bandwidth. When comparing the X700 boards to the other boards in the test we can see the theoretical effects of what the curious mix of performance elements gives them in comparison.
The X700 XT has large fill-rate and memory bandwidth deficits to the X800 XT, whilst having only a small triangle deficit as they both contain the same number of vertex shaders. Due to the increased numbers of pixel pipelines, the X700 XT has a has much larger fill-rates and even larger geometry processing capabilities than the X600 XT, whilst the bandwidth delta is less, but still reasonably large. In comparison to the 9800 PRO the fill-rate difference is smaller, going only by the difference clockspeeds, the but has a 23% bandwidth deficit.
Looking at the X700 PRO we see that there are still large theoretical rate differences between it and the X600 XT, although the bandwidth difference is much slimmer. Looking at the the X700 PRO in relation to the 9800 PRO, there is only a 12% fill-rate advantage for the X700 PRO, whilst the bandwidth deficit goes up to 37%.
Lets put these theoretical performance rates to the test using 3DMark2001SE's theoretical tests:
3DMark2001SE | Pixel Fill (Mpp/s) | Texel Fill (Mtp/s) | Geometry 1 Light (M tris/s) | Geometry 8 Lights (M tris/s) |
X700 XT | 1912.6 | 3704.4 | 161.5 | 34.9 |
X700 PRO | 1560.3 | 3332.0 | 148.3 | 31.3 |
X800 XT | 3347.1 | 7715.8 | 181.1 | 37.4 |
X600 XT | 1239.3 | 1973.0 | 71.8 | 12.7 |
9800 PRO | 2010.0 | 2954.5 | 108.0 | 19.1 |
X700 XT % Faster Than: | Pixel Fill | Texel Fill | Geometry 1 Light | Geometry 8 Lights |
X700 PRO | 22.6% | 11.2% | 8.9% | 11.5% |
X800 XT | -42.9% | -52.0% | -10.8% | -6.7% |
X600 XT | 54.3% | 87.8% | 124.9% | 174.8% |
9800 PRO | -4.8% | 25.4% | 49.5% | 82.7% |
X700 PRO % Faster Than: | Pixel Fill | Texel Fill | Geometry 1 Light | Geometry 8 Lights |
X800 XT | -53.4% | -56.8% | -18.1% | -16.3% |
X600 XT | 25.9% | 68.9% | 106.5% | 146.5% |
9800 PRO | -22.4% | 12.8% | 37.3% | 63.9% |
Predictably, the single texturing fill-rate performance difference between the X700 XT and X700 PRO falls inline with their bandwidth differences, whilst the multi-texturing test and geometry tests go, roughly, according to the theoretical differences. In fact, taking the peculiarities of the single texture fill-rate test here, all the relative performances are roughly where the theoretical differences between the boards would suggest they would be.