ATI RADEON 3850/3870 (RV670)
320 Shader Processors and 256-bit Memory Bus
Direct3D 9: Geometry Processing Speed Tests
Let's analyze extreme geometry tests. The first test uses the simplest vertex shader that shows maximum triangle throughput:
As all the GPUs are based on unified architectures, all unified processors in this test are busy with geometry processing. So all solutions demonstrate high results, which are evidently not limited by peak performance of unified processors, but by performance of other units, for example, triangle setup.
Test execution efficiency of various GPUs in various modes is approximately the same, peak performance in FFP, VS 1.1 and VS 2.0 modes is little different. Even these results show that the AMD solutions process geometry faster than NVIDIA GPUs. Performance of the RV670 matches that of the R600 adjusted for different frequencies. The RV630 is so strong probably owing to its higher clock rate. Let's see what will change in a more complex test with a single diffuse light source:
The situation did not change much, AMD solutions have an apparently higher potential. The GeForce 8800 GT lags further behind all AMD cards, including even the RADEON HD 2600 XT. The HD 3850 catches up with the HD 2900 XT. This time the FFP mode is a tad faster on two graphics cards: the GeForce 8800 and the HD 2900. Let's see what will happen in heavier conditions - complex lighting with a single light source and glares:
It's a similar situation. The apparent leader in geometry performance is a faster card based on the RV670. It's clear already that it succeeds not only owing to its clock rate. Something must be changed in the RV670, so it demonstrates better results in synthetic geometry tests than the R600.
This assumption is backed up by the fact that optimized FFP emulation appears with a mixed light source on most tested cards, except for the HD 3850 and the 3870. This time the GeForce 8800 GT is outperformed as much as in the previous case. Let's analyze the most complex geometry task with three light sources, including static and dynamic branches:
Here we can see differences between all contenders, these tests illustrate geometry performance well. Both new Mid-End solutions including the RADEON HD 2900 XT shoot forward. Perhaps, their potential is not revealed fully even in our most complex geometry task.
The HD 3850 and the HD 3870 act exactly like the HD 2900 XT. We are only surprised that the lower-clocked HD 3850 performs on a par with the HD 2900 XT. Perhaps, the reasons for this behavior may be in new architectural optimizations, larger caches, etc. In other cases, when shader units act as the main bottleneck, results of the cards line up in compliance with theoretical values. The RADEON HD 3870 is faster than its competitors, its results look especially good compared to the older Mid-End solution from AMD - the HD 2600 XT.
A brief conclusion on geometry tests: as the new GPUs are based on the good unified architecture R6xx, they perform great in geometry tests. They can use all unified stream processors to solve geometry tasks, and they are optimized for geometry processing. These solutions can even outperform the RADEON HD 2900 XT from time to time. However, unified processors will be busy mostly with pixels in real applications. We proceed to such tests now.
Your link here
